Full Code of axolotl-ai-cloud/axolotl for AI

gitextract_1sp7sr39/

├── .axolotl-complete.bash
├── .bandit
├── .coderabbit.yaml
├── .coveragerc
├── .editorconfig
├── .gitattributes
├── .github/
│   ├── CODE_OF_CONDUCT.md
│   ├── CONTRIBUTING.md
│   ├── FUNDING.yml
│   ├── ISSUE_TEMPLATE/
│   │   ├── bug-report.yaml
│   │   ├── config.yml
│   │   ├── docs.yml
│   │   └── feature-request.yaml
│   ├── PULL_REQUEST_TEMPLATE.md
│   ├── SECURITY.md
│   ├── SUPPORT.md
│   ├── release-drafter.yml
│   └── workflows/
│       ├── base.yml
│       ├── docs.yml
│       ├── lint.yml
│       ├── main.yml
│       ├── multi-gpu-e2e.yml
│       ├── nightlies.yml
│       ├── precommit-autoupdate.yml
│       ├── preview-docs.yml
│       ├── pypi.yml
│       ├── tests-nightly.yml
│       └── tests.yml
├── .gitignore
├── .mypy.ini
├── .pre-commit-config.yaml
├── .runpod/
│   ├── .gitignore
│   ├── Dockerfile
│   ├── README.md
│   ├── hub.json
│   ├── requirements.txt
│   ├── src/
│   │   ├── config/
│   │   │   └── config.yaml
│   │   ├── handler.py
│   │   ├── test_input.json
│   │   ├── train.py
│   │   └── utils.py
│   ├── test-input.json
│   └── tests.json
├── CITATION.cff
├── CNAME
├── FAQS.md
├── LICENSE
├── MANIFEST.in
├── README.md
├── VERSION
├── _quarto.yml
├── benchmarks/
│   ├── bench_entropy.py
│   ├── bench_scattermoe_lora.py
│   └── bench_selective_logsoftmax.py
├── cicd/
│   ├── Dockerfile-uv.jinja
│   ├── Dockerfile.jinja
│   ├── __init__.py
│   ├── cicd.sh
│   ├── cleanup.py
│   ├── cleanup.sh
│   ├── e2e_tests.py
│   ├── multigpu.py
│   ├── multigpu.sh
│   └── single_gpu.py
├── codecov.yml
├── deepspeed_configs/
│   ├── zero1.json
│   ├── zero1_torch_compile.json
│   ├── zero2.json
│   ├── zero2_torch_compile.json
│   ├── zero3.json
│   ├── zero3_bf16.json
│   ├── zero3_bf16_cpuoffload_all.json
│   └── zero3_bf16_cpuoffload_params.json
├── devtools/
│   ├── README.md
│   └── dev_chat_template.yml
├── docker/
│   ├── Dockerfile
│   ├── Dockerfile-base
│   ├── Dockerfile-base-next
│   ├── Dockerfile-base-nightly
│   ├── Dockerfile-cloud
│   ├── Dockerfile-cloud-no-tmux
│   ├── Dockerfile-cloud-uv
│   ├── Dockerfile-tests
│   ├── Dockerfile-uv
│   └── Dockerfile-uv-base
├── docker-compose.yaml
├── docs/
│   ├── .gitignore
│   ├── amd_hpc.qmd
│   ├── attention.qmd
│   ├── batch_vs_grad.qmd
│   ├── checkpoint_saving.qmd
│   ├── cli.qmd
│   ├── custom_integrations.qmd
│   ├── dataset-formats/
│   │   ├── conversation.qmd
│   │   ├── index.qmd
│   │   ├── inst_tune.qmd
│   │   ├── pretraining.qmd
│   │   ├── stepwise_supervised.qmd
│   │   ├── template_free.qmd
│   │   └── tokenized.qmd
│   ├── dataset_loading.qmd
│   ├── dataset_preprocessing.qmd
│   ├── debugging.qmd
│   ├── docker.qmd
│   ├── expert_quantization.qmd
│   ├── faq.qmd
│   ├── fsdp_qlora.qmd
│   ├── getting-started.qmd
│   ├── gradient_checkpointing.qmd
│   ├── inference.qmd
│   ├── input_output.qmd
│   ├── installation.qmd
│   ├── lora_optims.qmd
│   ├── lr_groups.qmd
│   ├── mac.qmd
│   ├── mixed_precision.qmd
│   ├── multi-gpu.qmd
│   ├── multi-node.qmd
│   ├── multimodal.qmd
│   ├── multipack.qmd
│   ├── nccl.qmd
│   ├── nd_parallelism.qmd
│   ├── optimizations.qmd
│   ├── optimizers.qmd
│   ├── qat.qmd
│   ├── quantize.qmd
│   ├── ray-integration.qmd
│   ├── reward_modelling.qmd
│   ├── rlhf.qmd
│   ├── scripts/
│   │   ├── examples-allowlist.yml
│   │   ├── generate_config_docs.py
│   │   └── generate_examples_docs.py
│   ├── sequence_parallelism.qmd
│   ├── streaming.qmd
│   ├── telemetry.qmd
│   ├── torchao.qmd
│   └── unsloth.qmd
├── examples/
│   ├── LiquidAI/
│   │   ├── README.md
│   │   ├── lfm2-350m-fft.yaml
│   │   ├── lfm2-8b-a1b-lora.yaml
│   │   └── lfm2-vl-lora.yaml
│   ├── alst/
│   │   ├── README.md
│   │   ├── llama3-8b-deepspeed-alst.yaml
│   │   └── llama3-8b-fsdp2-alst.yaml
│   ├── apertus/
│   │   ├── README.md
│   │   └── apertus-8b-qlora.yaml
│   ├── arcee/
│   │   ├── README.md
│   │   └── afm-4.5b-qlora.yaml
│   ├── archived/
│   │   ├── README.md
│   │   ├── cerebras/
│   │   │   ├── btlm-ft.yml
│   │   │   └── qlora.yml
│   │   ├── code-llama/
│   │   │   ├── 13b/
│   │   │   │   ├── lora.yml
│   │   │   │   └── qlora.yml
│   │   │   ├── 34b/
│   │   │   │   ├── lora.yml
│   │   │   │   └── qlora.yml
│   │   │   ├── 7b/
│   │   │   │   ├── lora.yml
│   │   │   │   └── qlora.yml
│   │   │   └── README.md
│   │   ├── dbrx/
│   │   │   ├── 16bit-lora.yaml
│   │   │   ├── 8bit-lora.yaml
│   │   │   ├── README.md
│   │   │   └── fft-ds-zero3.yaml
│   │   ├── deepcoder/
│   │   │   └── deepcoder-14B-preview-lora.yml
│   │   ├── falcon/
│   │   │   ├── config-7b-lora.yml
│   │   │   ├── config-7b-qlora.yml
│   │   │   └── config-7b.yml
│   │   ├── gemma/
│   │   │   └── qlora.yml
│   │   ├── gptj/
│   │   │   └── qlora.yml
│   │   ├── jeopardy-bot/
│   │   │   └── config.yml
│   │   ├── mpt-7b/
│   │   │   ├── README.md
│   │   │   └── config.yml
│   │   ├── openllama-3b/
│   │   │   ├── README.md
│   │   │   ├── config.yml
│   │   │   ├── lora.yml
│   │   │   └── qlora.yml
│   │   ├── pythia/
│   │   │   └── lora.yml
│   │   ├── pythia-12b/
│   │   │   ├── README.md
│   │   │   └── config.yml
│   │   ├── qwen/
│   │   │   ├── README.md
│   │   │   ├── lora.yml
│   │   │   ├── qlora.yml
│   │   │   ├── qwen2-moe-lora.yaml
│   │   │   └── qwen2-moe-qlora.yaml
│   │   ├── redpajama/
│   │   │   ├── README.md
│   │   │   └── config-3b.yml
│   │   ├── replit-3b/
│   │   │   └── config-lora.yml
│   │   ├── stablelm-2/
│   │   │   ├── 1.6b/
│   │   │   │   ├── fft.yml
│   │   │   │   └── lora.yml
│   │   │   └── README.md
│   │   ├── starcoder2/
│   │   │   └── qlora.yml
│   │   ├── tiny-llama/
│   │   │   ├── README.md
│   │   │   ├── lora-mps.yml
│   │   │   ├── lora.yml
│   │   │   ├── pretrain.yml
│   │   │   └── qlora.yml
│   │   ├── xgen-7b/
│   │   │   └── xgen-7b-8k-qlora.yml
│   │   └── yi-34B-chat/
│   │       ├── README.md
│   │       └── qlora.yml
│   ├── cloud/
│   │   ├── baseten.yaml
│   │   └── modal.yaml
│   ├── cohere/
│   │   └── command-r-7b-qlora.yml
│   ├── colab-notebooks/
│   │   └── colab-axolotl-example.ipynb
│   ├── deepcogito/
│   │   ├── cogito-v1-preview-llama-3B-lora.yml
│   │   └── cogito-v1-preview-qwen-14B-lora.yml
│   ├── deepseek-v2/
│   │   ├── fft-fsdp-16b.yaml
│   │   └── qlora-fsdp-2_5.yaml
│   ├── devstral/
│   │   ├── README.md
│   │   └── devstral-small-qlora.yml
│   ├── distributed-parallel/
│   │   ├── README.md
│   │   ├── llama-3_1-8b-hsdp-tp.yaml
│   │   └── qwen3-8b-fsdp-tp-cp.yaml
│   ├── eaft/
│   │   └── eaft-example.yml
│   ├── falcon-h1/
│   │   ├── falcon-h1-1b-deep-qlora.yaml
│   │   ├── falcon-h1-1b-qlora.yaml
│   │   ├── falcon-h1-34b-qlora.yaml
│   │   ├── falcon-h1-3b-qlora.yaml
│   │   ├── falcon-h1-500m-qlora.yaml
│   │   └── falcon-h1-7b-qlora.yaml
│   ├── gemma2/
│   │   ├── qlora.yml
│   │   └── reward-model.yaml
│   ├── gemma3/
│   │   ├── gemma-3-1b-qlora.yml
│   │   ├── gemma-3-270m-qlora.yml
│   │   ├── gemma-3-4b-qlora.yml
│   │   └── gemma-3-4b-vision-qlora.yml
│   ├── gemma3n/
│   │   ├── README.md
│   │   ├── gemma-3n-e2b-qlora.yml
│   │   ├── gemma-3n-e2b-vision-audio-qlora.yml
│   │   └── gemma-3n-e2b-vision-qlora.yml
│   ├── glm4/
│   │   └── qlora-32b.yaml
│   ├── glm45/
│   │   ├── README.md
│   │   └── glm-45-air-qlora.yaml
│   ├── glm46v/
│   │   ├── README.md
│   │   ├── glm-4-6v-flash-ddp.yaml
│   │   └── glm-4-6v-flash-qlora.yaml
│   ├── glm47-flash/
│   │   ├── README.md
│   │   ├── lora.yaml
│   │   ├── lora_fsdp.yaml
│   │   ├── qlora.yaml
│   │   └── qlora_fsdp.yaml
│   ├── gpt-oss/
│   │   ├── README.md
│   │   ├── gpt-oss-120b-fft-fsdp2-offload.yaml
│   │   ├── gpt-oss-20b-fft-deepspeed-zero3.yaml
│   │   ├── gpt-oss-20b-fft-fsdp2-offload.yaml
│   │   ├── gpt-oss-20b-fft-fsdp2.yaml
│   │   ├── gpt-oss-20b-sft-lora-singlegpu.yaml
│   │   └── gpt-oss-safeguard-20b-sft-lora-singlegpu.yaml
│   ├── granite4/
│   │   ├── README.md
│   │   └── granite-4.0-tiny-fft.yaml
│   ├── hunyuan/
│   │   ├── README.md
│   │   └── hunyuan-v1-dense-qlora.yaml
│   ├── internvl3_5/
│   │   ├── README.md
│   │   └── internvl3_5-8b-qlora.yml
│   ├── jamba/
│   │   ├── README.md
│   │   ├── qlora.yaml
│   │   ├── qlora_deepspeed.yaml
│   │   └── qlora_fsdp_large.yaml
│   ├── kimi-linear/
│   │   ├── README.md
│   │   └── kimi-48b-lora.yaml
│   ├── llama-2/
│   │   ├── README.md
│   │   ├── fft_optimized.yml
│   │   ├── gptq-lora.yml
│   │   ├── lisa.yml
│   │   ├── loftq.yml
│   │   ├── lora.yml
│   │   ├── qlora-fsdp.yml
│   │   ├── qlora.yml
│   │   └── relora.yml
│   ├── llama-3/
│   │   ├── 3b-fp8-fsdp2.yaml
│   │   ├── 3b-qat-fsdp2.yaml
│   │   ├── 3b-qat-mxfp4.yaml
│   │   ├── 3b-qat-nvfp4.yaml
│   │   ├── README.md
│   │   ├── diffusion/
│   │   │   ├── pretrain-1b.yaml
│   │   │   └── sft-1b.yaml
│   │   ├── fft-8b-liger-fsdp.yaml
│   │   ├── fft-8b.yaml
│   │   ├── instruct-dpo-lora-8b.yml
│   │   ├── instruct-lora-8b.yml
│   │   ├── lora-1b-deduplicate-dpo.yml
│   │   ├── lora-1b-deduplicate-sft.yml
│   │   ├── lora-1b-kernels.yml
│   │   ├── lora-1b-ray.yml
│   │   ├── lora-1b-sample-packing-sequentially.yml
│   │   ├── lora-1b.yml
│   │   ├── lora-8b.yml
│   │   ├── opentelemetry-qlora.yml
│   │   ├── qlora-1b-gdpo.yaml
│   │   ├── qlora-1b-kto.yaml
│   │   ├── qlora-1b.yml
│   │   ├── qlora-fsdp-405b.yaml
│   │   ├── qlora-fsdp-70b.yaml
│   │   ├── qlora.yml
│   │   └── sparse-finetuning.yaml
│   ├── llama-3-vision/
│   │   └── lora-11b.yaml
│   ├── llama-4/
│   │   ├── README.md
│   │   ├── do-no-use-fa2/
│   │   │   ├── maverick-qlora-fsdp1.yaml
│   │   │   ├── scout-qlora-fsdp1.yaml
│   │   │   ├── scout-qlora-single-h100.yaml
│   │   │   └── scout-vision-qlora-fsdp.yaml
│   │   ├── scout-qlora-flexattn-fsdp2.yaml
│   │   ├── scout-qlora-single-h100-flex.yaml
│   │   └── scout-vision-qlora-fsdp2-flex.yaml
│   ├── llava/
│   │   └── lora-7b.yaml
│   ├── magistral/
│   │   ├── README.md
│   │   ├── magistral-small-fsdp-qlora.yaml
│   │   ├── magistral-small-qlora.yaml
│   │   ├── think/
│   │   │   ├── README.md
│   │   │   └── magistral-small-think-qlora.yaml
│   │   └── vision/
│   │       ├── README.md
│   │       └── magistral-small-vision-24B-qlora.yml
│   ├── mamba/
│   │   └── config.yml
│   ├── mimo/
│   │   ├── README.md
│   │   └── mimo-7b-qlora.yaml
│   ├── ministral/
│   │   ├── README.md
│   │   └── ministral-small-qlora.yaml
│   ├── ministral3/
│   │   ├── README.md
│   │   ├── ministral3-3b-qlora.yaml
│   │   ├── think/
│   │   │   ├── README.md
│   │   │   └── ministral3-3b-think-qlora.yaml
│   │   └── vision/
│   │       ├── README.md
│   │       └── ministral3-3b-vision-qlora.yml
│   ├── mistral/
│   │   ├── README.md
│   │   ├── bigstral/
│   │   │   └── bigstral-ds-zero3.yaml
│   │   ├── config.yml
│   │   ├── dpo/
│   │   │   └── mistral-dpo-qlora.yml
│   │   ├── lora.yml
│   │   ├── mistral-qlora-fsdp.yml
│   │   ├── mixtral/
│   │   │   ├── mixtral-8x22b-qlora-fsdp.yml
│   │   │   ├── mixtral-qlora-fsdp.yml
│   │   │   ├── mixtral.yml
│   │   │   └── mixtral_22.yml
│   │   ├── mps/
│   │   │   └── lora-mps.yml
│   │   ├── orpo/
│   │   │   └── mistral-qlora-orpo.yml
│   │   └── qlora.yml
│   ├── mistral-small/
│   │   ├── README.md
│   │   └── mistral-small-3.1-24B-lora.yml
│   ├── mistral4/
│   │   ├── README.md
│   │   ├── fft-text.yml
│   │   ├── fft-vision.yml
│   │   ├── qlora-text.yml
│   │   └── qlora-vision.yml
│   ├── nemotron/
│   │   └── nemotron-mini-4b-qlora.yaml
│   ├── olmo3/
│   │   ├── README.md
│   │   └── olmo3-7b-qlora.yaml
│   ├── orpheus/
│   │   ├── README.md
│   │   └── finetune.yml
│   ├── phi/
│   │   ├── README.md
│   │   ├── lora-3.5.yaml
│   │   ├── phi-ft.yml
│   │   ├── phi-qlora.yml
│   │   ├── phi2-ft.yml
│   │   ├── phi3-ft-fsdp.yml
│   │   └── phi3-ft.yml
│   ├── pixtral/
│   │   └── lora-12b.yml
│   ├── plano/
│   │   ├── README.md
│   │   └── plano-4b-qlora.yaml
│   ├── qat_nvfp4/
│   │   ├── Gemma3-12B_baseline.yml
│   │   ├── Gemma3-12B_qat.yml
│   │   ├── Math-Gemma3-12B_baseline.yml
│   │   ├── Math-Gemma3-12B_qat.yml
│   │   ├── Math-Gemma3-27B_baseline.yml
│   │   ├── Math-Gemma3-27B_qat.yml
│   │   ├── Math-Qwen2.5-72B_baseline.yml
│   │   ├── Math-Qwen2.5-72B_qat.yml
│   │   ├── Qwen2.5-72B_baseline.yml
│   │   └── Qwen2.5-72B_qat.yml
│   ├── qwen2/
│   │   ├── adamw-pretrain-fsdp2.yaml
│   │   ├── dpo.yaml
│   │   ├── muon-pretrain-fsdp2.yaml
│   │   ├── prm.yaml
│   │   ├── qlora-fsdp.yaml
│   │   └── reward-model.yaml
│   ├── qwen2-vl/
│   │   └── lora-7b.yaml
│   ├── qwen2_5-vl/
│   │   └── lora-7b.yaml
│   ├── qwen3/
│   │   ├── 32b-qlora.yaml
│   │   ├── 8b-qat-fsdp2.yml
│   │   ├── README.md
│   │   ├── qlora-fsdp.yaml
│   │   └── reward-model.yaml
│   ├── qwen3-next/
│   │   ├── README.md
│   │   └── qwen3-next-80b-a3b-qlora.yaml
│   ├── qwen3.5/
│   │   ├── 122b-a10b-moe-qlora-fsdp.yaml
│   │   ├── 122b-a10b-moe-qlora.yaml
│   │   ├── 27b-fft.yaml
│   │   ├── 27b-qlora-fsdp.yaml
│   │   ├── 27b-qlora.yaml
│   │   ├── 35b-a3b-moe-qlora-fsdp.yaml
│   │   ├── 35b-a3b-moe-qlora.yaml
│   │   ├── 9b-fft-vision.yaml
│   │   ├── 9b-lora-vision.yaml
│   │   └── README.md
│   ├── seed-oss/
│   │   ├── README.md
│   │   └── seed-oss-36b-qlora.yaml
│   ├── slurm/
│   │   ├── README.md
│   │   └── axolotl.slurm
│   ├── smolvlm2/
│   │   ├── README.md
│   │   └── smolvlm2-2B-lora.yaml
│   ├── streaming/
│   │   ├── README.md
│   │   ├── pretrain.yaml
│   │   └── sft.yaml
│   ├── swanlab/
│   │   ├── README.md
│   │   ├── custom_trainer_profiling.py
│   │   ├── dpo-swanlab-completions.yml
│   │   ├── dpo-swanlab-full-featured.yml
│   │   └── lora-swanlab-profiling.yml
│   ├── trinity/
│   │   ├── README.md
│   │   └── trinity-nano-preview-qlora.yaml
│   └── voxtral/
│       ├── README.md
│       ├── voxtral-mini-audio-qlora.yml
│       └── voxtral-mini-qlora.yml
├── index.qmd
├── pyproject.toml
├── requirements-dev.txt
├── requirements-tests.txt
├── requirements.txt
├── scripts/
│   ├── chat_datasets.py
│   ├── cloud-entrypoint-term.sh
│   ├── cloud-entrypoint.sh
│   ├── cutcrossentropy_install.py
│   ├── motd
│   └── unsloth_install.py
├── setup.py
├── src/
│   ├── axolotl/
│   │   ├── __init__.py
│   │   ├── cli/
│   │   │   ├── __init__.py
│   │   │   ├── args.py
│   │   │   ├── art.py
│   │   │   ├── checks.py
│   │   │   ├── cloud/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── base.py
│   │   │   │   ├── baseten/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── template/
│   │   │   │   │       ├── run.sh
│   │   │   │   │       └── train_sft.py
│   │   │   │   └── modal_.py
│   │   │   ├── config.py
│   │   │   ├── delinearize_llama4.py
│   │   │   ├── evaluate.py
│   │   │   ├── inference.py
│   │   │   ├── main.py
│   │   │   ├── merge_lora.py
│   │   │   ├── merge_sharded_fsdp_weights.py
│   │   │   ├── preprocess.py
│   │   │   ├── quantize.py
│   │   │   ├── train.py
│   │   │   ├── utils/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── diffusion.py
│   │   │   │   ├── fetch.py
│   │   │   │   ├── load.py
│   │   │   │   ├── sweeps.py
│   │   │   │   └── train.py
│   │   │   └── vllm_serve.py
│   │   ├── common/
│   │   │   ├── __init__.py
│   │   │   ├── architectures.py
│   │   │   ├── const.py
│   │   │   └── datasets.py
│   │   ├── convert.py
│   │   ├── core/
│   │   │   ├── __init__.py
│   │   │   ├── attention/
│   │   │   │   └── __init__.py
│   │   │   ├── builders/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── base.py
│   │   │   │   ├── causal.py
│   │   │   │   └── rl.py
│   │   │   ├── chat/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── format/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── chatml.py
│   │   │   │   │   ├── llama3x.py
│   │   │   │   │   └── shared.py
│   │   │   │   └── messages.py
│   │   │   ├── datasets/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── chat.py
│   │   │   │   └── transforms/
│   │   │   │       ├── __init__.py
│   │   │   │       └── chat_builder.py
│   │   │   ├── trainers/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── base.py
│   │   │   │   ├── dpo/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── args.py
│   │   │   │   │   └── trainer.py
│   │   │   │   ├── grpo/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── args.py
│   │   │   │   │   ├── async_trainer.py
│   │   │   │   │   ├── fast_async_trainer.py
│   │   │   │   │   ├── replay_buffer.py
│   │   │   │   │   ├── sampler.py
│   │   │   │   │   └── trainer.py
│   │   │   │   ├── mamba.py
│   │   │   │   ├── mixins/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── activation_checkpointing.py
│   │   │   │   │   ├── checkpoints.py
│   │   │   │   │   ├── distributed_parallel.py
│   │   │   │   │   ├── optimizer.py
│   │   │   │   │   ├── packing.py
│   │   │   │   │   ├── rng_state_loader.py
│   │   │   │   │   └── scheduler.py
│   │   │   │   ├── trl.py
│   │   │   │   └── utils.py
│   │   │   ├── training_args.py
│   │   │   └── training_args_base.py
│   │   ├── datasets.py
│   │   ├── evaluate.py
│   │   ├── integrations/
│   │   │   ├── LICENSE.md
│   │   │   ├── __init__.py
│   │   │   ├── base.py
│   │   │   ├── config.py
│   │   │   ├── cut_cross_entropy/
│   │   │   │   ├── ACKNOWLEDGEMENTS.md
│   │   │   │   ├── LICENSE
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   └── args.py
│   │   │   ├── densemixer/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   └── plugin.py
│   │   │   ├── diffusion/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── callbacks.py
│   │   │   │   ├── generation.py
│   │   │   │   ├── plugin.py
│   │   │   │   ├── trainer.py
│   │   │   │   └── utils.py
│   │   │   ├── grokfast/
│   │   │   │   ├── LICENSE
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   └── optimizer.py
│   │   │   ├── kd/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── callbacks.py
│   │   │   │   ├── chat_template.py
│   │   │   │   ├── collator.py
│   │   │   │   ├── collator_online_teacher.py
│   │   │   │   ├── kernels/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── liger.py
│   │   │   │   │   └── models.py
│   │   │   │   ├── topk_logprob/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── forward_kl.py
│   │   │   │   ├── trainer.py
│   │   │   │   └── utils.py
│   │   │   ├── kernels/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── autotune_callback.py
│   │   │   │   ├── autotune_collector.py
│   │   │   │   ├── constants.py
│   │   │   │   ├── libs/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── scattermoe_lora/
│   │   │   │   │       ├── __init__.py
│   │   │   │   │       ├── kernels/
│   │   │   │   │       │   ├── __init__.py
│   │   │   │   │       │   ├── lora_ops.py
│   │   │   │   │       │   ├── ops.py
│   │   │   │   │       │   └── single.py
│   │   │   │   │       ├── layers.py
│   │   │   │   │       ├── lora_ops.py
│   │   │   │   │       ├── parallel_experts.py
│   │   │   │   │       ├── parallel_linear_lora.py
│   │   │   │   │       ├── selective_dequant.py
│   │   │   │   │       └── selective_dequant_kernel.py
│   │   │   │   ├── plugin.py
│   │   │   │   └── sonicmoe/
│   │   │   │       ├── __init__.py
│   │   │   │       ├── patch.py
│   │   │   │       ├── routing.py
│   │   │   │       └── weight_converter.py
│   │   │   ├── liger/
│   │   │   │   ├── LICENSE
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── models/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── base.py
│   │   │   │   │   ├── deepseekv2.py
│   │   │   │   │   ├── jamba.py
│   │   │   │   │   ├── llama4.py
│   │   │   │   │   ├── qwen3.py
│   │   │   │   │   └── qwen3_moe.py
│   │   │   │   ├── plugin.py
│   │   │   │   └── utils.py
│   │   │   ├── llm_compressor/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   ├── plugin.py
│   │   │   │   └── utils.py
│   │   │   ├── lm_eval/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   └── cli.py
│   │   │   ├── spectrum/
│   │   │   │   ├── LICENSE
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── args.py
│   │   │   │   └── model_snr_results/
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-1.5B-Instruct.json
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-1.5B.json
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-3B-Instruct.json
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-3B.json
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-7B-Instruct.json
│   │   │   │       ├── snr_results_Qwen-Qwen2.5-7B.json
│   │   │   │       ├── snr_results_google-gemma-2-2b.json
│   │   │   │       ├── snr_results_meta-llama-Llama-3.2-1B-Instruct.json
│   │   │   │       ├── snr_results_meta-llama-Llama-3.2-1B.json
│   │   │   │       ├── snr_results_meta-llama-Llama-3.2-3B-Instruct.json
│   │   │   │       └── snr_results_meta-llama-Llama-3.2-3B.json
│   │   │   └── swanlab/
│   │   │       ├── README.md
│   │   │       ├── __init__.py
│   │   │       ├── args.py
│   │   │       ├── callbacks.py
│   │   │       ├── completion_logger.py
│   │   │       ├── plugins.py
│   │   │       └── profiling.py
│   │   ├── kernels/
│   │   │   ├── __init__.py
│   │   │   ├── geglu.py
│   │   │   ├── lora.py
│   │   │   ├── quantize.py
│   │   │   ├── swiglu.py
│   │   │   └── utils.py
│   │   ├── loaders/
│   │   │   ├── __init__.py
│   │   │   ├── adapter.py
│   │   │   ├── adapters/
│   │   │   │   └── __init__.py
│   │   │   ├── constants.py
│   │   │   ├── model.py
│   │   │   ├── patch_manager.py
│   │   │   ├── processor.py
│   │   │   ├── tokenizer.py
│   │   │   └── utils.py
│   │   ├── logging_config.py
│   │   ├── models/
│   │   │   ├── __init__.py
│   │   │   └── mamba/
│   │   │       ├── __init__.py
│   │   │       ├── configuration_mamba.py
│   │   │       └── modeling_mamba.py
│   │   ├── monkeypatch/
│   │   │   ├── __init__.py
│   │   │   ├── accelerate/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── fsdp2.py
│   │   │   │   └── parallelism_config.py
│   │   │   ├── attention/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── flash_attn_4.py
│   │   │   │   ├── flex_attn.py
│   │   │   │   ├── sage_attn.py
│   │   │   │   └── xformers.py
│   │   │   ├── btlm_attn_hijack_flash.py
│   │   │   ├── data/
│   │   │   │   ├── __init__.py
│   │   │   │   └── batch_dataset_fetcher.py
│   │   │   ├── deepspeed_utils.py
│   │   │   ├── fsdp2_qlora.py
│   │   │   ├── gradient_checkpointing/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── offload_cpu.py
│   │   │   │   └── offload_disk.py
│   │   │   ├── llama_attn_hijack_flash.py
│   │   │   ├── llama_attn_hijack_xformers.py
│   │   │   ├── lora_kernels.py
│   │   │   ├── loss/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── chunked.py
│   │   │   │   └── eaft.py
│   │   │   ├── mistral_attn_hijack_flash.py
│   │   │   ├── mixtral/
│   │   │   │   └── __init__.py
│   │   │   ├── models/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── apertus/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── activation.py
│   │   │   │   ├── kimi_linear/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   ├── configuration_kimi.py
│   │   │   │   │   ├── modeling_kimi.py
│   │   │   │   │   ├── patch_kimi_linear.py
│   │   │   │   │   └── tokenization_kimi.py
│   │   │   │   ├── llama4/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── modeling.py
│   │   │   │   ├── mistral3/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── mistral_common_tokenizer.py
│   │   │   │   ├── pixtral/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── modeling_flash_attention_utils.py
│   │   │   │   ├── qwen3_5/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── modeling.py
│   │   │   │   ├── qwen3_next/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── modeling.py
│   │   │   │   └── voxtral/
│   │   │   │       ├── __init__.py
│   │   │   │       └── modeling.py
│   │   │   ├── moe_quant.py
│   │   │   ├── multipack.py
│   │   │   ├── peft/
│   │   │   │   ├── __init__.py
│   │   │   │   └── utils.py
│   │   │   ├── relora.py
│   │   │   ├── ring_attn/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── adapters/
│   │   │   │   │   ├── __init__.py
│   │   │   │   │   └── batch.py
│   │   │   │   └── patch.py
│   │   │   ├── scaled_softmax_attn.py
│   │   │   ├── stablelm_attn_hijack_flash.py
│   │   │   ├── tiled_mlp/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── base.py
│   │   │   │   └── patch.py
│   │   │   ├── trainer/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── lr.py
│   │   │   │   ├── trl.py
│   │   │   │   ├── trl_vllm.py
│   │   │   │   └── utils.py
│   │   │   ├── trainer_accelerator_args.py
│   │   │   ├── trainer_fsdp_optim.py
│   │   │   ├── transformers/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── trainer_context_parallel.py
│   │   │   │   └── trainer_loss_calc.py
│   │   │   ├── transformers_fa_utils.py
│   │   │   ├── unsloth_.py
│   │   │   ├── utils.py
│   │   │   └── xformers_/
│   │   │       └── __init__.py
│   │   ├── processing_strategies.py
│   │   ├── prompt_strategies/
│   │   │   ├── __init__.py
│   │   │   ├── alpaca_chat.py
│   │   │   ├── alpaca_instruct.py
│   │   │   ├── alpaca_w_system.py
│   │   │   ├── base.py
│   │   │   ├── bradley_terry/
│   │   │   │   ├── README.md
│   │   │   │   ├── __init__.py
│   │   │   │   ├── chat_template.py
│   │   │   │   └── llama3.py
│   │   │   ├── chat_template.py
│   │   │   ├── completion.py
│   │   │   ├── context_qa.py
│   │   │   ├── creative_acr.py
│   │   │   ├── dpo/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── chat_template.py
│   │   │   │   ├── chatml.py
│   │   │   │   ├── llama3.py
│   │   │   │   ├── passthrough.py
│   │   │   │   ├── user_defined.py
│   │   │   │   └── zephyr.py
│   │   │   ├── input_output.py
│   │   │   ├── jinja_template_analyzer.py
│   │   │   ├── kto/
│   │   │   │   ├── __init__.py
│   │   │   │   ├── chatml.py
│   │   │   │   ├── llama3.py
│   │   │   │   └── user_defined.py
│   │   │   ├── llama2_chat.py
│   │   │   ├── messages/
│   │   │   │   ├── __init__.py
│   │   │   │   └── chat.py
│   │   │   ├── metharme.py
│   │   │   ├── orcamini.py
│   │   │   ├── orpo/
│   │   │   │   ├── __init__.py
│   │   │   │   └── chat_template.py
│   │   │   ├── pretrain.py
│   │   │   ├── pygmalion.py
│   │   │   ├── stepwise_supervised.py
│   │   │   └── user_defined.py
│   │   ├── prompt_tokenizers.py
│   │   ├── prompters.py
│   │   ├── scripts/
│   │   │   ├── __init__.py
│   │   │   ├── vllm_serve_lora.py
│   │   │   └── vllm_worker_ext.py
│   │   ├── telemetry/
│   │   │   ├── __init__.py
│   │   │   ├── callbacks.py
│   │   │   ├── errors.py
│   │   │   ├── manager.py
│   │   │   ├── runtime_metrics.py
│   │   │   └── whitelist.yaml
│   │   ├── train.py
│   │   └── utils/
│   │       ├── __init__.py
│   │       ├── bench.py
│   │       ├── callbacks/
│   │       │   ├── __init__.py
│   │       │   ├── comet_.py
│   │       │   ├── dynamic_checkpoint.py
│   │       │   ├── generation.py
│   │       │   ├── lisa.py
│   │       │   ├── mlflow_.py
│   │       │   ├── models.py
│   │       │   ├── opentelemetry.py
│   │       │   ├── perplexity.py
│   │       │   ├── profiler.py
│   │       │   ├── qat.py
│   │       │   ├── swanlab.py
│   │       │   ├── tokens_per_second.py
│   │       │   └── trackio_.py
│   │       ├── chat_templates/
│   │       │   ├── __init__.py
│   │       │   ├── base.py
│   │       │   └── templates/
│   │       │       ├── alpaca.jinja
│   │       │       ├── aya.jinja
│   │       │       ├── chatml.jinja
│   │       │       ├── cohere.jinja
│   │       │       ├── command_a.jinja
│   │       │       ├── command_a_rag.jinja
│   │       │       ├── command_a_tool_use.jinja
│   │       │       ├── deepseek_v2.jinja
│   │       │       ├── deepseek_v3.jinja
│   │       │       ├── exaone.jinja
│   │       │       ├── exaone4.jinja
│   │       │       ├── falcon_h1.jinja
│   │       │       ├── gemma.jinja
│   │       │       ├── gemma3.jinja
│   │       │       ├── gemma3n.jinja
│   │       │       ├── jamba.jinja
│   │       │       ├── llama3.jinja
│   │       │       ├── llama3_2_vision.jinja
│   │       │       ├── llama4.jinja
│   │       │       ├── llava.jinja
│   │       │       ├── metharme.jinja
│   │       │       ├── mistral_v1.jinja
│   │       │       ├── mistral_v2v3.jinja
│   │       │       ├── mistral_v3_tekken.jinja
│   │       │       ├── mistral_v7_tekken.jinja
│   │       │       ├── phi_3.jinja
│   │       │       ├── phi_35.jinja
│   │       │       ├── phi_4.jinja
│   │       │       ├── pixtral.jinja
│   │       │       ├── qwen2_vl.jinja
│   │       │       ├── qwen3.jinja
│   │       │       ├── qwen3_5.jinja
│   │       │       └── qwen_25.jinja
│   │       ├── collators/
│   │       │   ├── __init__.py
│   │       │   ├── batching.py
│   │       │   ├── core.py
│   │       │   ├── mamba.py
│   │       │   └── mm_chat.py
│   │       ├── comet_.py
│   │       ├── config/
│   │       │   ├── __init__.py
│   │       │   └── models/
│   │       │       └── __init__.py
│   │       ├── ctx_managers/
│   │       │   ├── __init__.py
│   │       │   └── sequence_parallel.py
│   │       ├── data/
│   │       │   ├── __init__.py
│   │       │   ├── lock.py
│   │       │   ├── rl.py
│   │       │   ├── sft.py
│   │       │   ├── shared.py
│   │       │   ├── streaming.py
│   │       │   ├── utils.py
│   │       │   └── wrappers.py
│   │       ├── datasets.py
│   │       ├── dict.py
│   │       ├── distributed.py
│   │       ├── environment.py
│   │       ├── freeze.py
│   │       ├── generation/
│   │       │   ├── __init__.py
│   │       │   └── sft.py
│   │       ├── import_helper.py
│   │       ├── logging.py
│   │       ├── lora.py
│   │       ├── mistral/
│   │       │   ├── __init__.py
│   │       │   ├── mistral3_processor.py
│   │       │   └── mistral_tokenizer.py
│   │       ├── mlflow_.py
│   │       ├── model_shard_quant.py
│   │       ├── optimizers/
│   │       │   ├── __init__.py
│   │       │   └── adopt.py
│   │       ├── quantization.py
│   │       ├── samplers/
│   │       │   ├── __init__.py
│   │       │   ├── multipack.py
│   │       │   └── utils.py
│   │       ├── schedulers.py
│   │       ├── schemas/
│   │       │   ├── __init__.py
│   │       │   ├── config.py
│   │       │   ├── datasets.py
│   │       │   ├── deprecated.py
│   │       │   ├── dynamic_checkpoint.py
│   │       │   ├── enums.py
│   │       │   ├── fsdp.py
│   │       │   ├── integrations.py
│   │       │   ├── internal/
│   │       │   │   └── __init__.py
│   │       │   ├── model.py
│   │       │   ├── multimodal.py
│   │       │   ├── peft.py
│   │       │   ├── quantization.py
│   │       │   ├── training.py
│   │       │   ├── trl.py
│   │       │   ├── utils.py
│   │       │   ├── validation.py
│   │       │   └── vllm.py
│   │       ├── tee.py
│   │       ├── tokenization.py
│   │       ├── trackio_.py
│   │       ├── train.py
│   │       ├── trainer.py
│   │       └── wandb_.py
│   └── setuptools_axolotl_dynamic_dependencies.py
├── styles.css
└── tests/
    ├── __init__.py
    ├── cli/
    │   ├── __init__.py
    │   ├── conftest.py
    │   ├── test_cli_base.py
    │   ├── test_cli_evaluate.py
    │   ├── test_cli_fetch.py
    │   ├── test_cli_inference.py
    │   ├── test_cli_interface.py
    │   ├── test_cli_merge_lora.py
    │   ├── test_cli_merge_sharded_fsdp_weights.py
    │   ├── test_cli_preprocess.py
    │   ├── test_cli_sweeps.py
    │   ├── test_cli_train.py
    │   ├── test_cli_version.py
    │   ├── test_nested_options.py
    │   └── test_utils.py
    ├── conftest.py
    ├── constants.py
    ├── core/
    │   ├── chat/
    │   │   ├── __init__.py
    │   │   ├── format/
    │   │   │   └── __init__.py
    │   │   └── test_messages.py
    │   ├── test_async_grpo.py
    │   └── test_builders.py
    ├── e2e/
    │   ├── .gitignore
    │   ├── __init__.py
    │   ├── integrations/
    │   │   ├── test_cut_cross_entropy.py
    │   │   ├── test_fp8.py
    │   │   ├── test_hooks.py
    │   │   ├── test_kd.py
    │   │   ├── test_liger.py
    │   │   ├── test_llm_compressor.py
    │   │   ├── test_scattermoe_lora_kernels.py
    │   │   ├── test_scattermoe_lora_olmoe.py
    │   │   └── test_sonicmoe.py
    │   ├── kernels/
    │   │   ├── test_geglu.py
    │   │   ├── test_lora.py
    │   │   ├── test_quantize.py
    │   │   └── test_swiglu.py
    │   ├── multigpu/
    │   │   ├── __init__.py
    │   │   ├── patched/
    │   │   │   ├── __init__.py
    │   │   │   └── test_sp.py
    │   │   ├── solo/
    │   │   │   ├── __init__.py
    │   │   │   ├── test_flex.py
    │   │   │   ├── test_gdpo.py
    │   │   │   └── test_grpo.py
    │   │   ├── test_dist_muon_fsdp2.py
    │   │   ├── test_eval.py
    │   │   ├── test_fp8_fsdp2.py
    │   │   ├── test_fsdp1.py
    │   │   ├── test_fsdp2.py
    │   │   ├── test_gemma3.py
    │   │   ├── test_llama.py
    │   │   ├── test_locking.py
    │   │   ├── test_ray.py
    │   │   └── test_tp.py
    │   ├── patched/
    │   │   ├── __init__.py
    │   │   ├── lora_kernels/
    │   │   │   ├── __init__.py
    │   │   │   └── test_lora_kernel_patching.py
    │   │   ├── test_4d_multipack_llama.py
    │   │   ├── test_activation_checkpointing.py
    │   │   ├── test_cli_integrations.py
    │   │   ├── test_fa_xentropy.py
    │   │   ├── test_falcon_samplepack.py
    │   │   ├── test_flattening.py
    │   │   ├── test_fsdp2_qlora.py
    │   │   ├── test_fused_llama.py
    │   │   ├── test_llama_s2_attention.py
    │   │   ├── test_lora_llama_multipack.py
    │   │   ├── test_mistral_samplepack.py
    │   │   ├── test_mixtral_samplepack.py
    │   │   ├── test_model_patches.py
    │   │   ├── test_peft_embeddings.py
    │   │   ├── test_phi_multipack.py
    │   │   ├── test_resume.py
    │   │   ├── test_unsloth_integration.py
    │   │   └── test_unsloth_qlora.py
    │   ├── solo/
    │   │   ├── __init__.py
    │   │   ├── test_flex.py
    │   │   └── test_relora_llama.py
    │   ├── test_activation_offloading.py
    │   ├── test_deepseekv3.py
    │   ├── test_diffusion.py
    │   ├── test_dpo.py
    │   ├── test_embeddings_lr.py
    │   ├── test_evaluate.py
    │   ├── test_falcon.py
    │   ├── test_gemma2.py
    │   ├── test_gemma3_text.py
    │   ├── test_imports.py
    │   ├── test_llama.py
    │   ├── test_llama_pretrain.py
    │   ├── test_llama_vision.py
    │   ├── test_load_model.py
    │   ├── test_lora_llama.py
    │   ├── test_mamba.py
    │   ├── test_mistral.py
    │   ├── test_mixtral.py
    │   ├── test_optimizers.py
    │   ├── test_packing_loss.py
    │   ├── test_phi.py
    │   ├── test_preprocess.py
    │   ├── test_process_reward_model_smollm2.py
    │   ├── test_profiler.py
    │   ├── test_qat.py
    │   ├── test_quantization.py
    │   ├── test_qwen.py
    │   ├── test_reward_model_smollm2.py
    │   ├── test_save_first_step.py
    │   ├── test_schedulers.py
    │   ├── test_streaming.py
    │   ├── test_tokenizer.py
    │   └── utils.py
    ├── fixtures/
    │   ├── alpaca/
    │   │   └── alpaca.json
    │   ├── conversation.json
    │   ├── conversation.missingturns.json
    │   ├── conversation.tokenized.json
    │   └── conversation.tokenized_llama2chat.json
    ├── hf_offline_utils.py
    ├── integrations/
    │   ├── __init__.py
    │   ├── test_diffusion.py
    │   ├── test_diffusion_callback.py
    │   ├── test_kd_chat_template.py
    │   ├── test_liger.py
    │   ├── test_routing_parity.py
    │   ├── test_scattermoe_autotune_telemetry.py
    │   ├── test_scattermoe_lora.py
    │   ├── test_scattermoe_lora_kernels.py
    │   ├── test_sonicmoe.py
    │   ├── test_sonicmoe_gradients.py
    │   └── test_swanlab.py
    ├── monkeypatch/
    │   ├── test_llama_attn_hijack_flash.py
    │   ├── test_pixtral_flash_attention_patch.py
    │   ├── test_qwen3_next_modeling_patch.py
    │   ├── test_trainer_accelerator_args.py
    │   ├── test_trainer_context_parallel_patch.py
    │   ├── test_trainer_loss_calc.py
    │   ├── test_trl_vllm.py
    │   └── test_voxtral_modeling_patch.py
    ├── patched/
    │   └── test_validation.py
    ├── prompt_strategies/
    │   ├── __init__.py
    │   ├── conftest.py
    │   ├── messages/
    │   │   ├── __init__.py
    │   │   └── test_chat.py
    │   ├── test_alpaca.py
    │   ├── test_chat_template_ds_schema_unification.py
    │   ├── test_chat_template_utils.py
    │   ├── test_chat_templates.py
    │   ├── test_chat_templates_advanced.py
    │   ├── test_chat_templates_mistral.py
    │   ├── test_chat_templates_thinking.py
    │   ├── test_chat_templates_tool_call_string_arguments.py
    │   ├── test_dpo_chat_templates.py
    │   ├── test_dpo_chatml.py
    │   ├── test_jinja_template_analyzer.py
    │   ├── test_raw_io.py
    │   └── test_stepwise.py
    ├── telemetry/
    │   ├── __init__.py
    │   ├── conftest.py
    │   ├── test_callbacks.py
    │   ├── test_errors.py
    │   ├── test_manager.py
    │   └── test_runtime_metrics.py
    ├── test_chunked_xentropy.py
    ├── test_context_parallel_batch_size.py
    ├── test_convert.py
    ├── test_data.py
    ├── test_datasets.py
    ├── test_dict.py
    ├── test_exact_deduplication.py
    ├── test_freeze.py
    ├── test_loaders.py
    ├── test_logging_config_file_capture.py
    ├── test_lora.py
    ├── test_normalize_config.py
    ├── test_opentelemetry_callback.py
    ├── test_packed_batch_sampler.py
    ├── test_packed_dataset.py
    ├── test_packed_pretraining.py
    ├── test_perplexity.py
    ├── test_prompt_tokenizers.py
    ├── test_prompters.py
    ├── test_revision_parameter.py
    ├── test_save_deduplicated.py
    ├── test_schedulers.py
    ├── test_streaming.py
    ├── test_tensor_parallel_batch_size.py
    ├── test_tokenizers.py
    ├── test_train.py
    ├── test_triton_kernels.py
    ├── test_utils_tee.py
    ├── test_validation_dataset.py
    └── utils/
        ├── callbacks/
        │   └── test_dynamic_checkpoint.py
        ├── data/
        │   └── test_utils.py
        ├── lora/
        │   ├── test_config_validation_lora.py
        │   ├── test_freeze_lora.py
        │   └── test_merge_lora.py
        ├── schemas/
        │   └── validation/
        │       ├── test_activation_offloading.py
        │       ├── test_default_values.py
        │       ├── test_fsdp.py
        │       └── test_moe_quant.py
        ├── test_grpo_rw_fnc.py
        ├── test_import_helper.py
        ├── test_mistral3_processor.py
        └── test_train.py

SYMBOL INDEX (3886 symbols across 505 files)

FILE: .runpod/src/handler.py
  function handler (line 20) | async def handler(job):

FILE: .runpod/src/train.py
  function train (line 8) | async def train(config_path: str, gpu_id: str = "0", preprocess: bool = ...

FILE: .runpod/src/utils.py
  function get_output_dir (line 10) | def get_output_dir(run_id):
  function make_valid_config (line 15) | def make_valid_config(input_args):
  function set_config_env_vars (line 40) | def set_config_env_vars(args: dict):

FILE: benchmarks/bench_entropy.py
  function entropy_from_logits_original (line 20) | def entropy_from_logits_original(logits: torch.Tensor, chunk_size: int =...
  function _clean_gpu (line 33) | def _clean_gpu():
  function profile_time (line 41) | def profile_time(fn, logits, n_iters=BENCH_ITERS):
  function profile_memory (line 60) | def profile_memory(fn, logits, n_iters=MEM_ITERS):
  function fmt (line 77) | def fmt(values, unit=""):
  function benchmark_contiguous (line 83) | def benchmark_contiguous():
  function benchmark_noncontiguous (line 138) | def benchmark_noncontiguous():

FILE: benchmarks/bench_scattermoe_lora.py
  function _resolve_config (line 46) | def _resolve_config(spec):
  function _clean (line 79) | def _clean():
  function _bench (line 85) | def _bench(fn, warmup=WARMUP, iters=ITERS):
  function _setup (line 100) | def _setup(num_experts, K, N, T, top_k, R):
  function _call_fwd (line 117) | def _call_fwd(x, W, sei, ssi, top_k, lA, lB):
  function _call_base (line 130) | def _call_base(x, W, sei, ssi, top_k):
  function _call_dx (line 140) | def _call_dx(dy, W, sei, ssi, lA, lB):
  function _call_bwd (line 155) | def _call_bwd(dy, gx, lA, lB, eo, num_experts):
  function main (line 170) | def main():

FILE: benchmarks/bench_selective_logsoftmax.py
  function _clean_gpu (line 22) | def _clean_gpu():
  function profile_time (line 30) | def profile_time(fn, args, n_iters=BENCH_ITERS):
  function profile_memory (line 47) | def profile_memory(fn, args, n_iters=MEM_ITERS):
  function fmt (line 64) | def fmt(values, unit=""):
  function benchmark_forward (line 70) | def benchmark_forward():
  function benchmark_backward (line 123) | def benchmark_backward():

FILE: cicd/cleanup.py
  function cleanup (line 13) | def cleanup():
  function main (line 18) | def main():

FILE: cicd/e2e_tests.py
  function cicd_pytest (line 14) | def cicd_pytest():
  function main (line 19) | def main():

FILE: cicd/multigpu.py
  function run_cmd (line 63) | def run_cmd(cmd: str, run_folder: str):
  function cicd_pytest (line 79) | def cicd_pytest():
  function main (line 84) | def main():

FILE: cicd/single_gpu.py
  function run_cmd (line 64) | def run_cmd(cmd: str, run_folder: str):

FILE: docs/scripts/generate_config_docs.py
  class QuartoGenerator (line 20) | class QuartoGenerator:
    method __init__ (line 23) | def __init__(self):
    method _get_direct_fields (line 28) | def _get_direct_fields(self, cls: Type[BaseModel]) -> FrozenSet[str]:
    method _is_pydantic_model (line 46) | def _is_pydantic_model(self, type_obj) -> bool:
    method _extract_nested_type (line 50) | def _extract_nested_type(self, field_type) -> Any:
    method _extract_all_pydantic_models_from_type (line 126) | def _extract_all_pydantic_models_from_type(
    method _get_nested_models (line 186) | def _get_nested_models(
    method _build_inheritance_map (line 216) | def _build_inheritance_map(self, child_class: Type[BaseModel]):
    method _wrap_comment (line 237) | def _wrap_comment(self, text: str, width: int = 88) -> list[str]:
    method _extract_type_from_source (line 247) | def _extract_type_from_source(
    method _get_type_from_class_source (line 263) | def _get_type_from_class_source(self, class_obj: type, field_name: str...
    method _extract_field_groups_from_all_classes (line 285) | def _extract_field_groups_from_all_classes(
    method _extract_field_groups_from_source (line 319) | def _extract_field_groups_from_source(
    method _generate_field_documentation (line 438) | def _generate_field_documentation(
    method generate_qmd (line 591) | def generate_qmd(
  function main (line 736) | def main():

FILE: docs/scripts/generate_examples_docs.py
  function slugify (line 20) | def slugify(name: str) -> str:
  function read_allowlist (line 27) | def read_allowlist():
  function find_readme (line 36) | def find_readme(folder: Path) -> Path | None:
  function remove_first_h1 (line 44) | def remove_first_h1(md: str) -> tuple[str, str | None]:
  function rewrite_and_copy_assets (line 68) | def rewrite_and_copy_assets(md: str, src_dir: Path, dest_assets_root: Pa...
  function rewrite_readme_links (line 94) | def rewrite_readme_links(
  function write_qmd (line 209) | def write_qmd(out_path: Path, title: str, body_md: str):
  function update_quarto_yml (line 215) | def update_quarto_yml(generated: list[tuple[str, str, str]]):
  function main (line 286) | def main():

FILE: examples/swanlab/custom_trainer_profiling.py
  class CustomTrainerWithProfiling (line 30) | class CustomTrainerWithProfiling(AxolotlTrainer):
    method __init__ (line 39) | def __init__(self, *args, **kwargs):
    method training_step (line 56) | def training_step(self, model, inputs):
    method compute_loss (line 64) | def compute_loss(self, model, inputs, return_outputs=False):
    method prediction_step (line 72) | def prediction_step(self, model, inputs, prediction_loss_only, ignore_...
    method complex_training_step (line 85) | def complex_training_step(self, model, inputs):
    method _prepare_inputs (line 115) | def _prepare_inputs(self, inputs):
    method _prepare_input_for_model (line 130) | def _prepare_input_for_model(self, input_ids):
    method potentially_failing_method (line 147) | def potentially_failing_method(self):
    method _do_risky_computation (line 162) | def _do_risky_computation(self):
  class AdvancedProfilingTrainer (line 172) | class AdvancedProfilingTrainer(AxolotlTrainer):
    method __init__ (line 175) | def __init__(self, *args, **kwargs):
    method training_step (line 197) | def training_step(self, model, inputs):
    method _prepare_inputs (line 204) | def _prepare_inputs(self, inputs):
    method _debug_method (line 211) | def _debug_method(self, data):

FILE: scripts/chat_datasets.py
  function parse_dataset (line 13) | def parse_dataset(dataset=None, split="train"):

FILE: setup.py
  function parse_requirements (line 12) | def parse_requirements(extras_require_map):
  function get_package_version (line 147) | def get_package_version():

FILE: src/axolotl/cli/args.py
  class PreprocessCliArgs (line 8) | class PreprocessCliArgs:
  class TrainerCliArgs (line 29) | class TrainerCliArgs:
  class VllmServeCliArgs (line 40) | class VllmServeCliArgs:
  class QuantizeCliArgs (line 109) | class QuantizeCliArgs:
  class EvaluateCliArgs (line 122) | class EvaluateCliArgs:
  class InferenceCliArgs (line 131) | class InferenceCliArgs:

FILE: src/axolotl/cli/art.py
  function print_axolotl_text_art (line 22) | def print_axolotl_text_art():

FILE: src/axolotl/cli/checks.py
  function check_accelerate_default_config (line 16) | def check_accelerate_default_config() -> None:
  function check_user_token (line 24) | def check_user_token() -> bool:

FILE: src/axolotl/cli/cloud/__init__.py
  function load_cloud_cfg (line 16) | def load_cloud_cfg(cloud_config: Path | str) -> DictDefault:
  function do_cli_preprocess (line 24) | def do_cli_preprocess(
  function do_cli_train (line 35) | def do_cli_train(
  function do_cli_lm_eval (line 66) | def do_cli_lm_eval(

FILE: src/axolotl/cli/cloud/base.py
  class Cloud (line 9) | class Cloud(ABC):
    method preprocess (line 15) | def preprocess(self, config_yaml: str, *args, **kwargs) -> None:
    method train (line 19) | def train(

FILE: src/axolotl/cli/cloud/baseten/__init__.py
  class BasetenCloud (line 14) | class BasetenCloud(Cloud):
    method __init__ (line 17) | def __init__(self, config: dict):
    method preprocess (line 20) | def preprocess(self, config_yaml: str, *args, **kwargs) -> None:
    method train (line 26) | def train(

FILE: src/axolotl/cli/cloud/modal_.py
  function run_cmd (line 18) | def run_cmd(cmd: str, run_folder: str, volumes=None):
  class ModalCloud (line 52) | class ModalCloud(Cloud):
    method __init__ (line 57) | def __init__(self, config, app=None):
    method get_env (line 69) | def get_env(self):
    method get_image (line 84) | def get_image(self):
    method get_secrets (line 128) | def get_secrets(self):
    method create_volume (line 140) | def create_volume(self, volume_config):
    method get_ephemeral_disk_size (line 145) | def get_ephemeral_disk_size(self):
    method get_preprocess_timeout (line 148) | def get_preprocess_timeout(self):
    method get_preprocess_memory (line 153) | def get_preprocess_memory(self):
    method get_preprocess_env (line 161) | def get_preprocess_env(self):
    method preprocess (line 172) | def preprocess(self, config_yaml: str, *args, **kwargs):
    method get_train_timeout (line 183) | def get_train_timeout(self):
    method get_train_gpu (line 188) | def get_train_gpu(self):
    method get_train_memory (line 208) | def get_train_memory(self):
    method get_train_env (line 214) | def get_train_env(self, local_dirs=None):
    method train (line 228) | def train(
    method lm_eval (line 247) | def lm_eval(self, config_yaml: str):
  function _preprocess (line 261) | def _preprocess(config_yaml: str, volumes=None):
  function _train (line 273) | def _train(
  function _lm_eval (line 307) | def _lm_eval(config_yaml: str, volumes=None):

FILE: src/axolotl/cli/config.py
  function _coerce_value (line 36) | def _coerce_value(value: Any, existing: Optional[Any] = None) -> Any:
  function check_remote_config (line 97) | def check_remote_config(config: Union[str, Path]) -> Union[str, Path]:
  function choose_config (line 163) | def choose_config(path: Path) -> str:
  function prepare_plugins (line 208) | def prepare_plugins(cfg: DictDefault):
  function plugin_set_cfg (line 223) | def plugin_set_cfg(cfg: DictDefault):
  function load_cfg (line 230) | def load_cfg(
  function compute_supports_fp8 (line 349) | def compute_supports_fp8() -> bool:

FILE: src/axolotl/cli/delinearize_llama4.py
  function iter_convert_patched_to_hf (line 15) | def iter_convert_patched_to_hf(model_state_dict, num_experts) -> Generator:
  function do_cli (line 71) | def do_cli(model: Union[Path, str], output: Union[Path, str]) -> None:

FILE: src/axolotl/cli/evaluate.py
  function do_evaluate (line 21) | def do_evaluate(cfg: DictDefault, cli_args: TrainerCliArgs) -> None:
  function do_cli (line 44) | def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs) -> None:

FILE: src/axolotl/cli/inference.py
  function get_multi_line_input (line 32) | def get_multi_line_input() -> str:
  function do_inference (line 50) | def do_inference(
  function do_inference_gradio (line 168) | def do_inference_gradio(
  function do_cli (line 299) | def do_cli(

FILE: src/axolotl/cli/main.py
  function cli (line 43) | def cli():
  function preprocess (line 57) | def preprocess(config: str, cloud: Optional[str] = None, **kwargs):
  function train (line 98) | def train(
  function evaluate (line 153) | def evaluate(ctx: click.Context, config: str, launcher: str, **kwargs):
  function inference (line 198) | def inference(ctx: click.Context, config: str, launcher: str, gradio: bo...
  function merge_sharded_fsdp_weights (line 245) | def merge_sharded_fsdp_weights(
  function merge_lora (line 282) | def merge_lora(config: str, **kwargs):
  function fetch (line 299) | def fetch(directory: str, dest: Optional[str]):
  function vllm_serve (line 318) | def vllm_serve(config: str, **cli_args: VllmServeCliArgs):
  function quantize (line 328) | def quantize(config: str, **cli_args: QuantizeCliArgs):
  function delinearize_llama4 (line 337) | def delinearize_llama4(model: str, output: str):
  function main (line 346) | def main():

FILE: src/axolotl/cli/merge_lora.py
  function do_merge_lora (line 18) | def do_merge_lora(*, cfg: DictDefault) -> None:
  function do_cli (line 55) | def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs) -> None:

FILE: src/axolotl/cli/merge_sharded_fsdp_weights.py
  class BFloat16CastPlanner (line 31) | class BFloat16CastPlanner(_EmptyStateDictLoadPlanner):
    method commit_tensor (line 34) | def commit_tensor(self, read_item, tensor):
  function _distributed_checkpoint_to_merged_weights (line 38) | def _distributed_checkpoint_to_merged_weights(
  function merge_fsdp_weights (line 108) | def merge_fsdp_weights(
  function do_cli (line 169) | def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):

FILE: src/axolotl/cli/preprocess.py
  function do_preprocess (line 29) | def do_preprocess(cfg: DictDefault, cli_args: PreprocessCliArgs) -> None:
  function do_cli (line 99) | def do_cli(

FILE: src/axolotl/cli/quantize.py
  function do_quantize (line 23) | def do_quantize(

FILE: src/axolotl/cli/train.py
  function do_train (line 23) | def do_train(cfg: DictDefault, cli_args: TrainerCliArgs):
  function do_cli (line 55) | def do_cli(config: Union[Path, str] = Path("examples/"), **kwargs):
  function ray_train_func (line 91) | def ray_train_func(kwargs: dict):

FILE: src/axolotl/cli/utils/args.py
  function _strip_optional_type (line 12) | def _strip_optional_type(field_type: type | str | None):
  function filter_none_kwargs (line 32) | def filter_none_kwargs(func: Callable) -> Callable:
  function add_options_from_dataclass (line 53) | def add_options_from_dataclass(config_class: Type[Any]) -> Callable:
  function _is_pydantic_model (line 91) | def _is_pydantic_model(field_type: type) -> bool:
  function _get_field_description (line 99) | def _get_field_description(field) -> str | None:
  function _add_nested_model_options (line 108) | def _add_nested_model_options(
  function add_options_from_config (line 148) | def add_options_from_config(config_class: Type[BaseModel]) -> Callable:

FILE: src/axolotl/cli/utils/diffusion.py
  function diffusion_inference (line 12) | def diffusion_inference(
  function _parse_commands (line 91) | def _parse_commands(text: str):
  function run_diffusion (line 139) | def run_diffusion(
  function render_html (line 199) | def render_html(
  function launch_diffusion_gradio_ui (line 260) | def launch_diffusion_gradio_ui(

FILE: src/axolotl/cli/utils/fetch.py
  function _download_file (line 16) | def _download_file(
  function fetch_from_github (line 70) | def fetch_from_github(

FILE: src/axolotl/cli/utils/load.py
  function load_model_and_tokenizer (line 20) | def load_model_and_tokenizer(

FILE: src/axolotl/cli/utils/sweeps.py
  function generate_sweep_configs (line 9) | def generate_sweep_configs(

FILE: src/axolotl/cli/utils/train.py
  function _add_default_rdzv_args (line 15) | def _add_default_rdzv_args(launcher_args: list[str]) -> list[str]:
  function build_command (line 46) | def build_command(base_cmd: list[str], options: dict[str, Any]) -> list[...
  function generate_config_files (line 69) | def generate_config_files(config: str, sweep: str | None) -> Iterator[tu...
  function launch_training (line 109) | def launch_training(
  function _launch_cloud_training (line 134) | def _launch_cloud_training(
  function _launch_accelerate_training (line 157) | def _launch_accelerate_training(
  function _launch_torchrun_training (line 195) | def _launch_torchrun_training(
  function _launch_python_training (line 221) | def _launch_python_training(cfg_file: str, kwargs: dict) -> None:

FILE: src/axolotl/cli/vllm_serve.py
  class AxolotlScriptArguments (line 15) | class AxolotlScriptArguments(ScriptArguments):
  function do_vllm_serve (line 24) | def do_vllm_serve(

FILE: src/axolotl/common/datasets.py
  class TrainDatasetMeta (line 23) | class TrainDatasetMeta:
  function sample_dataset (line 31) | def sample_dataset(dataset: Dataset, num_samples: int) -> Dataset:
  function load_datasets (line 39) | def load_datasets(
  function load_preference_datasets (line 102) | def load_preference_datasets(

FILE: src/axolotl/convert.py
  class FileReader (line 7) | class FileReader:
    method read (line 12) | def read(self, file_path):
  class FileWriter (line 17) | class FileWriter:
    method __init__ (line 22) | def __init__(self, file_path):
    method write (line 25) | def write(self, content):
  class StdoutWriter (line 30) | class StdoutWriter:
    method write (line 35) | def write(self, content):
  class JsonParser (line 40) | class JsonParser:
    method parse (line 45) | def parse(self, content):
  class JsonlSerializer (line 49) | class JsonlSerializer:
    method serialize (line 54) | def serialize(self, data):
  class JsonToJsonlConverter (line 59) | class JsonToJsonlConverter:
    method __init__ (line 64) | def __init__(self, file_reader, file_writer, json_parser, jsonl_serial...
    method convert (line 70) | def convert(self, input_file_path):

FILE: src/axolotl/core/builders/base.py
  class TrainerBuilderBase (line 55) | class TrainerBuilderBase(abc.ABC):
    method __init__ (line 58) | def __init__(self, cfg, model, tokenizer, processor=None):
    method model_ref (line 78) | def model_ref(self):
    method model_ref (line 82) | def model_ref(self, model):
    method train_dataset (line 86) | def train_dataset(self):
    method train_dataset (line 90) | def train_dataset(self, dataset):
    method eval_dataset (line 94) | def eval_dataset(self):
    method eval_dataset (line 98) | def eval_dataset(self, dataset):
    method peft_config (line 102) | def peft_config(self):
    method peft_config (line 106) | def peft_config(self, peft_config):
    method build (line 110) | def build(self, total_num_steps):
    method get_callbacks (line 113) | def get_callbacks(self) -> list[TrainerCallback]:
    method get_post_trainer_create_callbacks (line 182) | def get_post_trainer_create_callbacks(self, trainer):
    method hook_pre_create_training_args (line 200) | def hook_pre_create_training_args(self, training_arguments_kwargs):
    method hook_post_create_training_args (line 204) | def hook_post_create_training_args(self, training_arguments):
    method hook_pre_create_trainer (line 208) | def hook_pre_create_trainer(self, trainer_kwargs, trainer_cls):
    method hook_post_create_trainer (line 212) | def hook_post_create_trainer(self, trainer):
    method _configure_warmup_and_logging (line 216) | def _configure_warmup_and_logging(
    method _configure_precision_settings (line 251) | def _configure_precision_settings(self, training_args_kwargs: dict):
    method _configure_scheduler (line 261) | def _configure_scheduler(self, training_args_kwargs: dict):
    method _configure_optimizer (line 273) | def _configure_optimizer(self, training_args_kwargs: dict, trainer_kwa...
    method _configure_hub_parameters (line 426) | def _configure_hub_parameters(self, training_args_kwargs: dict):
    method _configure_save_and_eval_strategy (line 439) | def _configure_save_and_eval_strategy(self, training_args_kwargs: dict):
    method _configure_reporting (line 466) | def _configure_reporting(self, training_args_kwargs: dict):
    method _configure_torch_compile (line 490) | def _configure_torch_compile(self, training_args_kwargs: dict):
    method _configure_accelerator_config (line 502) | def _configure_accelerator_config(self, training_args_kwargs: dict):
    method _configure_gradient_checkpointing (line 510) | def _configure_gradient_checkpointing(self, training_args_kwargs: dict):
    method _set_base_training_args (line 528) | def _set_base_training_args(

FILE: src/axolotl/core/builders/causal.py
  class HFCausalTrainerBuilder (line 53) | class HFCausalTrainerBuilder(TrainerBuilderBase):
    method get_callbacks (line 59) | def get_callbacks(self):
    method get_post_trainer_create_callbacks (line 82) | def get_post_trainer_create_callbacks(self, trainer):
    method _get_trainer_cls (line 134) | def _get_trainer_cls(self):
    method build (line 163) | def build(self, total_num_steps):
    method build_collator (line 454) | def build_collator(

FILE: src/axolotl/core/builders/rl.py
  class HFRLTrainerBuilder (line 24) | class HFRLTrainerBuilder(TrainerBuilderBase):
    method get_callbacks (line 27) | def get_callbacks(self):
    method get_post_trainer_create_callbacks (line 35) | def get_post_trainer_create_callbacks(self, trainer):
    method _get_trainer_cls (line 39) | def _get_trainer_cls(self, trainer_kwargs: dict):
    method _build_training_arguments (line 96) | def _build_training_arguments(self, total_num_steps):
    method build (line 206) | def build(self, total_num_steps):

FILE: src/axolotl/core/chat/format/chatml.py
  function format_message (line 11) | def format_message(

FILE: src/axolotl/core/chat/format/llama3x.py
  function format_message (line 11) | def format_message(message: Messages, message_index: Optional[int] = Non...

FILE: src/axolotl/core/chat/format/shared.py
  function wrap_tools (line 8) | def wrap_tools(message: Messages):

FILE: src/axolotl/core/chat/messages.py
  class MessageRoles (line 13) | class MessageRoles(str, Enum):
  class MessageContentTypes (line 28) | class MessageContentTypes(str, Enum):
  class SpecialToken (line 41) | class SpecialToken(str, Enum):
  class ToolCallFunction (line 50) | class ToolCallFunction(BaseModel):
  class Tool (line 59) | class Tool(BaseModel):
  class ToolCallContents (line 69) | class ToolCallContents(BaseModel):
    method __str__ (line 78) | def __str__(self) -> str:
  class ToolResponseContents (line 85) | class ToolResponseContents(BaseModel):
    method __str__ (line 94) | def __str__(self) -> str:
  class MessageContents (line 101) | class MessageContents(BaseModel):
    method __str__ (line 113) | def __str__(self) -> str:
  class Messages (line 120) | class Messages(BaseModel):
    method __str__ (line 131) | def __str__(self) -> str:
    method tokenized (line 134) | def tokenized(
  class Chats (line 180) | class Chats(BaseModel):
    method __str__ (line 187) | def __str__(self) -> str:
    method tokenized (line 190) | def tokenized(
  class ChatFormattedChats (line 208) | class ChatFormattedChats(Chats):
    method model_post_init (line 216) | def model_post_init(self, __context):
  class PreferenceChats (line 223) | class PreferenceChats(BaseModel):

FILE: src/axolotl/core/datasets/chat.py
  class TokenizedChatDataset (line 13) | class TokenizedChatDataset(Dataset):
    method __init__ (line 18) | def __init__(

FILE: src/axolotl/core/datasets/transforms/chat_builder.py
  function chat_message_transform_builder (line 9) | def chat_message_transform_builder(

FILE: src/axolotl/core/trainers/base.py
  class AxolotlTrainer (line 64) | class AxolotlTrainer(
    method axolotl_cfg (line 81) | def axolotl_cfg(self):
    method axolotl_cfg (line 85) | def axolotl_cfg(self, cfg):
    method __init__ (line 88) | def __init__(
    method _create_multipack_sampler (line 109) | def _create_multipack_sampler(
    method _get_train_sampler (line 150) | def _get_train_sampler(
    method _get_eval_sampler (line 187) | def _get_eval_sampler(self, eval_dataset: Dataset | None = None) -> Sa...
    method _get_dataloader (line 219) | def _get_dataloader(
    method _get_bench_sampler (line 316) | def _get_bench_sampler(
    method get_bench_dataloader (line 323) | def get_bench_dataloader(
    method compute_loss (line 344) | def compute_loss(
    method evaluate (line 399) | def evaluate(self, *args, **kwargs):
    method orpo_concatenate_inputs (line 404) | def orpo_concatenate_inputs(inputs, label_pad_token=-100, pad_token=0,...
    method orpo_compute_custom_loss (line 455) | def orpo_compute_custom_loss(self, logits, labels):
    method orpo_compute_logps (line 473) | def orpo_compute_logps(
    method orpo_compute_loss (line 497) | def orpo_compute_loss(
    method push_to_hub (line 565) | def push_to_hub(self, *args, **kwargs) -> str:
    method create_accelerator_and_postprocess (line 578) | def create_accelerator_and_postprocess(self):
    method additional_accelerator_args (line 587) | def additional_accelerator_args(
    method log (line 609) | def log(self, logs: dict[str, float], start_time: float | None = None)...
    method store_metrics (line 672) | def store_metrics(
    method _save_checkpoint (line 695) | def _save_checkpoint(self, model, trial, **kwargs):
    method _save (line 717) | def _save(self, output_dir: Optional[str] = None, state_dict=None):

FILE: src/axolotl/core/trainers/dpo/__init__.py
  class DPOStrategy (line 7) | class DPOStrategy:
    method get_trainer_class (line 11) | def get_trainer_class(cls):
    method get_training_args_class (line 15) | def get_training_args_class(cls):
    method set_training_args_kwargs (line 21) | def set_training_args_kwargs(cls, cfg):

FILE: src/axolotl/core/trainers/dpo/args.py
  class AxolotlDPOConfig (line 13) | class AxolotlDPOConfig(AxolotlTrainingMixins, DPOConfig):

FILE: src/axolotl/core/trainers/dpo/trainer.py
  class AxolotlDPOTrainer (line 23) | class AxolotlDPOTrainer(
    method __init__ (line 35) | def __init__(self, *args, dataset_tags=None, **kwargs):
    method push_to_hub (line 43) | def push_to_hub(self, *args, **kwargs) -> str:
    method tokenize_row (line 57) | def tokenize_row(
    method training_step (line 87) | def training_step(
    method concatenated_forward (line 98) | def concatenated_forward(

FILE: src/axolotl/core/trainers/grpo/__init__.py
  class GRPOStrategy (line 26) | class GRPOStrategy:
    method get_trainer_class (line 30) | def get_trainer_class(
    method get_training_args_class (line 51) | def get_training_args_class(
    method set_training_args_kwargs (line 59) | def set_training_args_kwargs(cls, cfg: DictDefault) -> dict[str, Any]:
    method set_trainer_args (line 205) | def set_trainer_args(cls, cfg: DictDefault) -> list[Any]:
    method set_trainer_kwargs (line 216) | def set_trainer_kwargs(cls, cfg: DictDefault) -> dict[str, Any]:
    method get_collator (line 228) | def get_collator(cls, *args, **kwargs):
    method get_blocklist_args_kwargs (line 233) | def get_blocklist_args_kwargs(cls) -> list[str]:
    method get_reward_func (line 242) | def get_reward_func(cls, reward_func_fqn: str) -> RewardFunc:
    method get_rollout_func (line 286) | def get_rollout_func(cls, rollout_func_fqn: str):

FILE: src/axolotl/core/trainers/grpo/args.py
  class AxolotlGRPOConfig (line 14) | class AxolotlGRPOConfig(AxolotlTrainingMixins, GRPOConfig):
  class AxolotlAsyncGRPOConfig (line 21) | class AxolotlAsyncGRPOConfig(AxolotlTrainingMixins, FastAsyncGRPOConfig):

FILE: src/axolotl/core/trainers/grpo/async_trainer.py
  function disable_gradient_checkpointing (line 68) | def disable_gradient_checkpointing(model, kwargs):
  class AsyncGRPOConfig (line 98) | class AsyncGRPOConfig(GRPOConfig):
  class ProducerConfig (line 184) | class ProducerConfig:
    method __post_init__ (line 211) | def __post_init__(self):
  class DataProducer (line 232) | class DataProducer(ABC):
    method produce (line 242) | def produce(
  class BaseDataProducer (line 256) | class BaseDataProducer(DataProducer):
    method __init__ (line 259) | def __init__(self, config: ProducerConfig | None = None):
    method on_rollout_begin (line 262) | def on_rollout_begin(self, global_step: int) -> None:
    method on_rollout_end (line 265) | def on_rollout_end(self, dataset: Dataset, global_step: int) -> None:
  class AsyncDataProducer (line 269) | class AsyncDataProducer:
    method __init__ (line 282) | def __init__(
    method config (line 303) | def config(self) -> ProducerConfig:
    method produce (line 306) | def produce(self, model: Any, global_step: int, **kwargs) -> Dataset:
    method _broadcast_dataset (line 358) | def _broadcast_dataset(self, dataset) -> Dataset:
    method _locked_produce (line 391) | def _locked_produce(self, model: Any, global_step: int, **kwargs) -> D...
    method on_rollout_begin (line 396) | def on_rollout_begin(self, global_step: int) -> None:
    method on_rollout_end (line 400) | def on_rollout_end(self, dataset: Dataset, global_step: int) -> None:
    method shutdown (line 404) | def shutdown(self) -> None:
  class DataProducerCallback (line 412) | class DataProducerCallback:
  class RolloutDataset (line 424) | class RolloutDataset(Dataset):
    method __init__ (line 434) | def __init__(self, data: dict[str, Any]):
    method __len__ (line 461) | def __len__(self) -> int:
    method __getitem__ (line 464) | def __getitem__(self, idx: int) -> dict[str, Any]:
  function make_rollout_collator (line 473) | def make_rollout_collator(shared_keys: set[str]):
  class GRPODataProducer (line 492) | class GRPODataProducer(BaseDataProducer):
    method __init__ (line 499) | def __init__(
    method set_trainer (line 523) | def set_trainer(self, trainer) -> None:
    method _init_prompt_dataloader (line 528) | def _init_prompt_dataloader(self) -> None:
    method produce (line 570) | def produce(
  class AsyncGRPOTrainer (line 623) | class AsyncGRPOTrainer(GRPOTrainer):
    method __init__ (line 630) | def __init__(self, *args, **kwargs):
    method _create_data_producer (line 709) | def _create_data_producer(self, args, train_dataset):
    method _setup_async (line 743) | def _setup_async(self):
    method _shutdown_async (line 777) | def _shutdown_async(self):
    method _submit_generation (line 782) | def _submit_generation(self):
    method _sync_peft_weights_no_merge (line 792) | def _sync_peft_weights_no_merge(self):
    method _sync_lora_adapter (line 892) | def _sync_lora_adapter(self):
    method _maybe_sync_vllm_weights (line 998) | def _maybe_sync_vllm_weights(self):
    method _zero_pad_embedding_for_fp8 (line 1045) | def _zero_pad_embedding_for_fp8(self):
    method _generate_single_turn (line 1091) | def _generate_single_turn(self, prompts, **kwargs):
    method _generate_rank0_only (line 1129) | def _generate_rank0_only(self, prompts):
    method _generate_only (line 1217) | def _generate_only(self, inputs, rank0_only=False):
    method _compute_rewards_for_batch (line 1414) | def _compute_rewards_for_batch(
    method _launch_reward_workers (line 1422) | def _launch_reward_workers(self, inputs, prompts, completions, complet...
    method _collect_reward_workers (line 1429) | def _collect_reward_workers(
    method _post_advantage_hook (line 1444) | def _post_advantage_hook(
    method _compute_deferred_scores (line 1463) | def _compute_deferred_scores(self, rollout: dict) -> dict:
    method _compute_streaming_group_scores (line 1802) | def _compute_streaming_group_scores(
    method _score_streaming (line 2153) | def _score_streaming(self, rollout: dict) -> list[dict]:
    method _prepare_inputs (line 2215) | def _prepare_inputs(self, generation_batch):
    method _prepare_inputs_data_producer (line 2230) | def _prepare_inputs_data_producer(self, generation_batch):
    method _prepare_inputs_legacy_async (line 2271) | def _prepare_inputs_legacy_async(self, generation_batch):
    method _get_per_token_logps_and_entropies (line 2302) | def _get_per_token_logps_and_entropies(
    method get_off_policy_mask (line 2437) | def get_off_policy_mask(
    method _compute_loss (line 2453) | def _compute_loss(self, model, inputs):

FILE: src/axolotl/core/trainers/grpo/fast_async_trainer.py
  class FastAsyncGRPOConfig (line 51) | class FastAsyncGRPOConfig(AsyncGRPOConfig):
  class RerollDataProducer (line 117) | class RerollDataProducer(GRPODataProducer):
    method _pre_produce_hook (line 125) | def _pre_produce_hook(self, inputs: list, global_step: int) -> list:
  function _persistent_reward_worker (line 169) | def _persistent_reward_worker(conn):
  class FastAsyncGRPOTrainer (line 216) | class FastAsyncGRPOTrainer(AsyncGRPOTrainer):
    method __init__ (line 226) | def __init__(self, *args, **kwargs):
    method _create_data_producer (line 265) | def _create_data_producer(self, args, train_dataset):
    method _get_reward_workers (line 301) | def _get_reward_workers(self):
    method _shutdown_reward_workers (line 328) | def _shutdown_reward_workers(self):
    method _compute_rewards_for_batch (line 346) | def _compute_rewards_for_batch(
    method _launch_reward_workers (line 355) | def _launch_reward_workers(self, inputs, prompts, completions, complet...
    method _collect_reward_workers (line 410) | def _collect_reward_workers(
    method _post_advantage_hook (line 470) | def _post_advantage_hook(
    method compute_liger_loss (line 637) | def compute_liger_loss(self, unwrapped_model, inputs):
    method _compute_loss (line 753) | def _compute_loss(self, model, inputs):

FILE: src/axolotl/core/trainers/grpo/replay_buffer.py
  class ReplayBuffer (line 8) | class ReplayBuffer:
    method __init__ (line 14) | def __init__(self, max_size: int):
    method __len__ (line 19) | def __len__(self):
    method add (line 22) | def add(self, score: float, data: dict):
    method sample (line 32) | def sample(self, num_samples: int) -> list[dict] | None:

FILE: src/axolotl/core/trainers/grpo/sampler.py
  class SequenceParallelRepeatRandomSampler (line 13) | class SequenceParallelRepeatRandomSampler(Sampler):
    method __init__ (line 54) | def __init__(
    method __iter__ (line 109) | def __iter__(self) -> Iterator[int]:
    method __len__ (line 155) | def __len__(self) -> int:
    method set_epoch (line 166) | def set_epoch(self, epoch: int) -> None:

FILE: src/axolotl/core/trainers/grpo/trainer.py
  class AxolotlGRPOTrainer (line 57) | class AxolotlGRPOTrainer(
  class AxolotlAsyncGRPOTrainer (line 70) | class AxolotlAsyncGRPOTrainer(
  class AxolotlGRPOSequenceParallelTrainer (line 83) | class AxolotlGRPOSequenceParallelTrainer(AxolotlGRPOTrainer):
    method __init__ (line 86) | def __init__(
    method train (line 166) | def train(self, *args, **kwargs):
    method _get_train_sampler (line 176) | def _get_train_sampler(self) -> Sampler:
    method _create_dataloader_params (line 198) | def _create_dataloader_params(self, is_eval=False, custom_batch_size=N...
    method _prepare_dataloader (line 221) | def _prepare_dataloader(
    method get_train_dataloader (line 264) | def get_train_dataloader(self) -> DataLoader:
    method _generate_and_score_completions (line 291) | def _generate_and_score_completions(

FILE: src/axolotl/core/trainers/mamba.py
  class AxolotlMambaTrainer (line 8) | class AxolotlMambaTrainer(AxolotlTrainer):
    method compute_loss (line 13) | def compute_loss(

FILE: src/axolotl/core/trainers/mixins/activation_checkpointing.py
  class ActivationOffloadingMixin (line 25) | class ActivationOffloadingMixin(Trainer):
    method __init__ (line 30) | def __init__(self, *args, **kwargs):
    method training_step (line 44) | def training_step(self, *args, **kwargs):
  function ac_wrap_hf_model (line 49) | def ac_wrap_hf_model(model: nn.Module, **kwargs):
  function get_lora_act_offloading_ctx_manager (line 54) | def get_lora_act_offloading_ctx_manager(

FILE: src/axolotl/core/trainers/mixins/checkpoints.py
  class CheckpointSaveMixin (line 10) | class CheckpointSaveMixin(Trainer):
    method _save_optimizer_and_scheduler (line 13) | def _save_optimizer_and_scheduler(self, output_dir):

FILE: src/axolotl/core/trainers/mixins/distributed_parallel.py
  class DistributedParallelMixin (line 9) | class DistributedParallelMixin(Trainer):
    method _save (line 14) | def _save(self, output_dir: str | None = None, state_dict=None):
    method create_accelerator_and_postprocess (line 23) | def create_accelerator_and_postprocess(self):

FILE: src/axolotl/core/trainers/mixins/optimizer.py
  class OptimizerMixin (line 17) | class OptimizerMixin(Trainer):
    method create_optimizer_grouped_parameters (line 22) | def create_optimizer_grouped_parameters(
    method create_optimizer (line 107) | def create_optimizer(self, model=None):
  class OptimizerInitMixin (line 201) | class OptimizerInitMixin:
    method __init__ (line 207) | def __init__(self, *args, **kwargs):

FILE: src/axolotl/core/trainers/mixins/packing.py
  class PackingMixin (line 6) | class PackingMixin(Trainer):
    method _set_signature_columns_if_needed (line 11) | def _set_signature_columns_if_needed(self):

FILE: src/axolotl/core/trainers/mixins/rng_state_loader.py
  class RngLoaderMixin (line 24) | class RngLoaderMixin(Trainer):
    method _load_rng_state (line 29) | def _load_rng_state(self, checkpoint):

FILE: src/axolotl/core/trainers/mixins/scheduler.py
  class SchedulerMixin (line 20) | class SchedulerMixin(Trainer):
    method create_scheduler (line 27) | def create_scheduler(

FILE: src/axolotl/core/trainers/trl.py
  class AxolotlORPOTrainer (line 14) | class AxolotlORPOTrainer(
  class AxolotlKTOTrainer (line 29) | class AxolotlKTOTrainer(
  class AxolotlCPOTrainer (line 44) | class AxolotlCPOTrainer(
  class AxolotlRewardTrainer (line 59) | class AxolotlRewardTrainer(
  class AxolotlPRMTrainer (line 74) | class AxolotlPRMTrainer(

FILE: src/axolotl/core/trainers/utils.py
  function sanitize_kwargs_for_tagging (line 4) | def sanitize_kwargs_for_tagging(tag_names, kwargs=None):
  function sanitize_kwargs_for_ds_tagging (line 20) | def sanitize_kwargs_for_ds_tagging(dataset_tags, kwargs=None):

FILE: src/axolotl/core/training_args.py
  class AxolotlTrainingArguments (line 23) | class AxolotlTrainingArguments(AxolotlTrainingMixins, TrainingArguments):
  class AxolotlORPOConfig (line 33) | class AxolotlORPOConfig(AxolotlTrainingMixins, ORPOConfig):
  class AxolotlKTOConfig (line 40) | class AxolotlKTOConfig(AxolotlTrainingMixins, KTOConfig):
  class AxolotlCPOConfig (line 47) | class AxolotlCPOConfig(AxolotlTrainingMixins, CPOConfig):
  class AxolotlRewardConfig (line 59) | class AxolotlRewardConfig(AxolotlTrainingMixins, RewardConfig):
  class AxolotlPRMConfig (line 66) | class AxolotlPRMConfig(AxolotlTrainingMixins, PRMConfig):

FILE: src/axolotl/core/training_args_base.py
  class AxolotlTrainingMixins (line 12) | class AxolotlTrainingMixins:

FILE: src/axolotl/datasets.py
  class TokenizedPromptDataset (line 18) | class TokenizedPromptDataset(Dataset):
    method __init__ (line 28) | def __init__(
    method process (line 44) | def process(self, dataset):
  function wrap_dataset_for_tokenized_prompt (line 72) | def wrap_dataset_for_tokenized_prompt(

FILE: src/axolotl/evaluate.py
  function evaluate_dataset (line 30) | def evaluate_dataset(
  function evaluate (line 68) | def evaluate(*, cfg: DictDefault, dataset_meta: TrainDatasetMeta) -> Dic...

FILE: src/axolotl/integrations/base.py
  class BasePlugin (line 44) | class BasePlugin:
    method __init__ (line 76) | def __init__(self):
    method register (line 79) | def register(self, cfg: dict):
    method get_input_args (line 86) | def get_input_args(self) -> str | None:
    method get_training_args_mixin (line 89) | def get_training_args_mixin(self) -> str | None:
    method load_datasets (line 94) | def load_datasets(
    method pre_model_load (line 107) | def pre_model_load(self, cfg: DictDefault):
    method post_model_build (line 114) | def post_model_build(self, cfg: DictDefault, model: PreTrainedModel):
    method pre_lora_load (line 121) | def pre_lora_load(self, cfg: DictDefault, model: PreTrainedModel):
    method post_lora_load (line 129) | def post_lora_load(self, cfg: DictDefault, model: PreTrainedModel | Pe...
    method post_model_load (line 137) | def post_model_load(self, cfg: DictDefault, model: PreTrainedModel | P...
    method get_trainer_cls (line 145) | def get_trainer_cls(self, cfg: DictDefault) -> type[Trainer] | None:
    method post_trainer_create (line 155) | def post_trainer_create(self, cfg: DictDefault, trainer: Trainer):
    method get_training_args (line 163) | def get_training_args(self, cfg: DictDefault):
    method get_collator_cls_and_kwargs (line 174) | def get_collator_cls_and_kwargs(self, cfg: DictDefault, is_eval: bool ...
    method create_optimizer (line 186) | def create_optimizer(self, cfg: DictDefault, trainer: Trainer) -> Opti...
    method create_lr_scheduler (line 197) | def create_lr_scheduler(
    method add_callbacks_pre_trainer (line 216) | def add_callbacks_pre_trainer(
    method add_callbacks_post_trainer (line 230) | def add_callbacks_post_trainer(
    method post_train (line 245) | def post_train(self, cfg: DictDefault, model: PreTrainedModel | PeftMo...
    method post_train_unload (line 253) | def post_train_unload(self, cfg: DictDefault):
  function load_plugin (line 261) | def load_plugin(plugin_name: str) -> BasePlugin:
  class PluginManager (line 301) | class PluginManager:
    method __new__ (line 320) | def __new__(cls):
    method get_instance (line 330) | def get_instance() -> "PluginManager":
    method cfg (line 339) | def cfg(self):
    method cfg (line 343) | def cfg(self, cfg):
    method register (line 346) | def register(self, plugin_name: str):
    method get_input_args (line 366) | def get_input_args(self) -> list[str]:
    method get_training_args_mixin (line 379) | def get_training_args_mixin(self):
    method load_datasets (line 393) | def load_datasets(
    method pre_model_load (line 415) | def pre_model_load(self, cfg: DictDefault):
    method post_model_build (line 424) | def post_model_build(self, cfg: DictDefault, model: PreTrainedModel):
    method pre_lora_load (line 435) | def pre_lora_load(self, cfg: DictDefault, model: PreTrainedModel):
    method post_lora_load (line 445) | def post_lora_load(self, cfg: DictDefault, model: PreTrainedModel | Pe...
    method post_model_load (line 455) | def post_model_load(self, cfg: DictDefault, model: PreTrainedModel | P...
    method get_trainer_cls (line 466) | def get_trainer_cls(self, cfg: DictDefault) -> Trainer | None:
    method get_training_args (line 482) | def get_training_args(self, cfg):
    method get_collator_cls_and_kwargs (line 500) | def get_collator_cls_and_kwargs(self, cfg, is_eval=False):
    method post_trainer_create (line 518) | def post_trainer_create(self, cfg: DictDefault, trainer: Trainer):
    method create_optimizer (line 528) | def create_optimizer(self, trainer: Trainer) -> Optimizer | None:
    method create_lr_scheduler (line 544) | def create_lr_scheduler(
    method add_callbacks_pre_trainer (line 568) | def add_callbacks_pre_trainer(
    method add_callbacks_post_trainer (line 587) | def add_callbacks_post_trainer(
    method post_train (line 606) | def post_train(self, cfg: DictDefault, model: PreTrainedModel | PeftMo...
    method post_train_unload (line 616) | def post_train_unload(self, cfg: DictDefault):
  class BaseOptimizerFactory (line 626) | class BaseOptimizerFactory:
    method __call__ (line 629) | def __call__(
    method get_decay_parameter_names (line 635) | def get_decay_parameter_names(self, model) -> list[str]:

FILE: src/axolotl/integrations/config.py
  function merge_input_args (line 27) | def merge_input_args():
  function merge_training_args (line 60) | def merge_training_args() -> Type:

FILE: src/axolotl/integrations/cut_cross_entropy/__init__.py
  class CutCrossEntropyPlugin (line 42) | class CutCrossEntropyPlugin(BasePlugin):
    method get_input_args (line 47) | def get_input_args(self):
    method _check_requirements (line 50) | def _check_requirements(self):
    method pre_model_load (line 86) | def pre_model_load(self, cfg):
    method patch_llama_like (line 105) | def patch_llama_like(

FILE: src/axolotl/integrations/cut_cross_entropy/args.py
  class CutCrossEntropyArgs (line 28) | class CutCrossEntropyArgs(BaseModel):
    method check_dtype_is_half (line 37) | def check_dtype_is_half(cls, data):
    method check_chunked_cross_entropy_not_set (line 48) | def check_chunked_cross_entropy_not_set(cls, data):

FILE: src/axolotl/integrations/densemixer/args.py
  class DenseMixerArgs (line 6) | class DenseMixerArgs(BaseModel):

FILE: src/axolotl/integrations/densemixer/plugin.py
  class DenseMixerPlugin (line 11) | class DenseMixerPlugin(BasePlugin):
    method get_input_args (line 16) | def get_input_args(self) -> str | None:
    method pre_model_load (line 19) | def pre_model_load(self, cfg):

FILE: src/axolotl/integrations/diffusion/args.py
  class DiffusionConfig (line 10) | class DiffusionConfig(BaseModel):
    method _validate_mask_ratios (line 83) | def _validate_mask_ratios(self) -> "DiffusionConfig":
  class DiffusionArgs (line 89) | class DiffusionArgs(BaseModel):

FILE: src/axolotl/integrations/diffusion/callbacks.py
  class DiffusionGenerationCallback (line 23) | class DiffusionGenerationCallback(TrainerCallback):
    method __init__ (line 26) | def __init__(self, trainer):
    method on_step_end (line 29) | def on_step_end(
    method _log_samples (line 70) | def _log_samples(self, samples: list, step: int):

FILE: src/axolotl/integrations/diffusion/generation.py
  function generate_samples (line 15) | def generate_samples(
  function _sample_sequences_from_dataloader (line 103) | def _sample_sequences_from_dataloader(
  function generate (line 196) | def generate(
  function _clean_masked_text (line 321) | def _clean_masked_text(masked_text: str, tokenizer: Any, mask_token_id: ...
  function _diffusion_step (line 339) | def _diffusion_step(

FILE: src/axolotl/integrations/diffusion/plugin.py
  class DiffusionPlugin (line 15) | class DiffusionPlugin(BasePlugin):
    method __init__ (line 23) | def __init__(self):
    method get_input_args (line 27) | def get_input_args(self) -> str:
    method post_model_load (line 31) | def post_model_load(self, cfg: DictDefault, model: PreTrainedModel | P...
    method get_trainer_cls (line 35) | def get_trainer_cls(self, cfg: DictDefault) -> type[DiffusionTrainer] ...
    method post_trainer_create (line 39) | def post_trainer_create(self, cfg: DictDefault, trainer: DiffusionTrai...

FILE: src/axolotl/integrations/diffusion/trainer.py
  class DiffusionTrainer (line 19) | class DiffusionTrainer(AxolotlTrainer):
    method __init__ (line 22) | def __init__(self, *args, **kwargs):
    method set_config (line 27) | def set_config(self, config: DictDefault):
    method _resolve_mask_token_id (line 40) | def _resolve_mask_token_id(self) -> None:
    method compute_loss (line 59) | def compute_loss(
    method _cache_special_token_ids (line 82) | def _cache_special_token_ids(self):
    method _forward_process (line 100) | def _forward_process(
    method _compute_diffusion_loss (line 159) | def _compute_diffusion_loss(

FILE: src/axolotl/integrations/diffusion/utils.py
  function resolve_mask_token_id (line 12) | def resolve_mask_token_id(
  function create_bidirectional_attention_mask (line 125) | def create_bidirectional_attention_mask(
  function shift_logits_to_input_positions (line 162) | def shift_logits_to_input_positions(logits: torch.Tensor) -> torch.Tensor:

FILE: src/axolotl/integrations/grokfast/__init__.py
  class GrokfastCallbackHandler (line 16) | class GrokfastCallbackHandler(TrainerCallback):
    method __init__ (line 21) | def __init__(self, *args_, alpha=0.98, lamb=2.0, **kwargs):
    method on_train_begin (line 27) | def on_train_begin(self, *args_, **kwargs):
    method on_pre_optimizer_step (line 30) | def on_pre_optimizer_step(self, args_, state, control, **kwargs):
  class GrokfastPlugin (line 36) | class GrokfastPlugin(BasePlugin):
    method get_input_args (line 41) | def get_input_args(self):
    method add_callbacks_post_trainer (line 44) | def add_callbacks_post_trainer(self, cfg, trainer):

FILE: src/axolotl/integrations/grokfast/args.py
  class GrokfastArgs (line 10) | class GrokfastArgs(BaseModel):

FILE: src/axolotl/integrations/grokfast/optimizer.py
  function gradfilter_ma (line 11) | def gradfilter_ma(
  function gradfilter_ema (line 44) | def gradfilter_ema(

FILE: src/axolotl/integrations/kd/__init__.py
  class KDPlugin (line 29) | class KDPlugin(BasePlugin):
    method get_input_args (line 34) | def get_input_args(self):
    method get_training_args_mixin (line 37) | def get_training_args_mixin(self):
    method get_trainer_cls (line 40) | def get_trainer_cls(self, cfg):
    method get_training_args (line 47) | def get_training_args(self, cfg):
    method get_collator_cls_and_kwargs (line 56) | def get_collator_cls_and_kwargs(self, cfg, is_eval=False):
    method pre_model_load (line 84) | def pre_model_load(self, cfg):
    method add_callbacks_post_trainer (line 89) | def add_callbacks_post_trainer(self, cfg: Any, trainer: Trainer) -> list:

FILE: src/axolotl/integrations/kd/args.py
  class InferenceServerType (line 25) | class InferenceServerType(str, Enum):
  class KDArgs (line 34) | class KDArgs(BaseModel):
  class KDTrainingArgsMixin (line 63) | class KDTrainingArgsMixin:

FILE: src/axolotl/integrations/kd/callbacks.py
  class KDTemperatureSchedulerCallback (line 10) | class KDTemperatureSchedulerCallback(TrainerCallback):
    method __init__ (line 15) | def __init__(self, temperature_start, temperature_min, trainer):
    method on_step_end (line 22) | def on_step_end(self, args, state, control, **kwargs):

FILE: src/axolotl/integrations/kd/chat_template.py
  class ChatTemplateStrategyWithKD (line 29) | class ChatTemplateStrategyWithKD(ChatTemplateStrategy):
    method __init__ (line 34) | def __init__(
    method supports_batched (line 66) | def supports_batched(self) -> bool:
    method transform_logprobs (line 70) | def transform_logprobs(self, sample):
    method _tokenize_single_prompt (line 178) | def _tokenize_single_prompt(self, prompt):
    method _prepare_kd_fields (line 189) | def _prepare_kd_fields(self, tokenized_prompt, original_prompt):
  class ChatTemplateStrategyWithKDv2 (line 196) | class ChatTemplateStrategyWithKDv2(ChatTemplateStrategyWithKD):
    method transform_logprobs (line 201) | def transform_logprobs(self, sample):
    method _prepare_kd_fields (line 295) | def _prepare_kd_fields(self, tokenized_prompt, original_prompt):
  class KDStrategyLoader (line 305) | class KDStrategyLoader(StrategyLoader):
    method _get_strategy_cls (line 310) | def _get_strategy_cls(self, cfg):
    method _get_strategy_params (line 313) | def _get_strategy_params(self, cfg, ds_cfg: Dict[str, Any]):
  class KDStrategyLoaderV2 (line 325) | class KDStrategyLoaderV2(KDStrategyLoader):
    method _get_strategy_cls (line 330) | def _get_strategy_cls(self, cfg):

FILE: src/axolotl/integrations/kd/collator.py
  class DataCollatorForKD (line 32) | class DataCollatorForKD(DataCollatorForSeq2Seq):
    method __init__ (line 49) | def __init__(self, *args, **kwargs):
    method __call__ (line 53) | def __call__(self, features, return_tensors=None):
  class KDBatchSamplerDataCollatorForSeq2Seq (line 198) | class KDBatchSamplerDataCollatorForSeq2Seq(DataCollatorForKD):
    method __call__ (line 204) | def __call__(self, features, return_tensors=None):

FILE: src/axolotl/integrations/kd/collator_online_teacher.py
  function hmac_sha_from_int_list (line 21) | def hmac_sha_from_int_list(int_list, key, hash_func=hashlib.sha256):
  class OnlineTeacherCollator (line 46) | class OnlineTeacherCollator(KDBatchSamplerDataCollatorForSeq2Seq):
    method __init__ (line 53) | def __init__(
    method _normalize_logprobs (line 85) | def _normalize_logprobs(self, raw_logprobs: List[float]) -> List[float]:
    method fetch_online_logprobs_sglang (line 98) | def fetch_online_logprobs_sglang(
    method fetch_online_logprobs_vllm (line 267) | def fetch_online_logprobs_vllm(
    method __call__ (line 467) | def __call__(

FILE: src/axolotl/integrations/kd/kernels/liger.py
  class LigerFusedLinearKLTopKLogprobFunction (line 14) | class LigerFusedLinearKLTopKLogprobFunction(LigerFusedLinearDistillation...
    method distillation_loss_fn (line 20) | def distillation_loss_fn(
    method _compute_loss_kl_topk (line 120) | def _compute_loss_kl_topk(
    method forward (line 180) | def forward(
    method backward (line 352) | def backward(ctx, grad_output):
  class LigerFusedLinearKLTopKLogprobLoss (line 415) | class LigerFusedLinearKLTopKLogprobLoss(torch.nn.Module):
    method __init__ (line 420) | def __init__(
    method forward (line 458) | def forward(

FILE: src/axolotl/integrations/kd/kernels/models.py
  class TransformersKwargs (line 15) | class TransformersKwargs(FlashAttentionKwargs, LossKwargs):
  function kldiv_forward_llama_like (line 28) | def kldiv_forward_llama_like(
  function apply_kernel (line 98) | def apply_kernel(model_type):

FILE: src/axolotl/integrations/kd/topk_logprob/forward_kl.py
  function loss (line 24) | def loss(
  class ChunkedTopKKDLoss (line 100) | class ChunkedTopKKDLoss(nn.Module):
    method __init__ (line 108) | def __init__(self, num_output_chunks: int = 8, kd_temperature: float =...
    method forward (line 113) | def forward(

FILE: src/axolotl/integrations/kd/trainer.py
  class AxolotlKDTrainer (line 26) | class AxolotlKDTrainer(AxolotlTrainer):
    method __init__ (line 31) | def __init__(self, *args, **kwargs):
    method _set_signature_columns_if_needed (line 52) | def _set_signature_columns_if_needed(self):
    method compute_loss (line 66) | def compute_loss(

FILE: src/axolotl/integrations/kd/utils.py
  function normalize_logprobs (line 11) | def normalize_logprobs(logprobs: FloatTensor, topk: int) -> FloatTensor:
  function strided_chunk_views (line 46) | def strided_chunk_views(
  function chunk_overlap (line 94) | def chunk_overlap(input_tensor: Tensor, chunks: int, dim: int = 0, overl...

FILE: src/axolotl/integrations/kernels/args.py
  class KernelsArgs (line 8) | class KernelsArgs(BaseModel):
    method check_mutually_exclusive (line 14) | def check_mutually_exclusive(cls, data):
    method check_use_kernels (line 24) | def check_use_kernels(cls, data):
    method check_experts_implementation (line 35) | def check_experts_implementation(cls, data):
    method disable_mlp_kernel (line 50) | def disable_mlp_kernel(cls, data):

FILE: src/axolotl/integrations/kernels/autotune_callback.py
  function _get_gpu_info (line 19) | def _get_gpu_info() -> dict:
  function _get_smem_capacity (line 35) | def _get_smem_capacity() -> dict:
  class AutotuneReportCallback (line 53) | class AutotuneReportCallback(TrainerCallback):
    method __init__ (line 67) | def __init__(self):
    method on_step_end (line 71) | def on_step_end(

FILE: src/axolotl/integrations/kernels/autotune_collector.py
  function _parse_key_tuple (line 28) | def _parse_key_tuple(key_tuple: tuple) -> dict[str, Any]:
  function _find_lora_ops_module (line 44) | def _find_lora_ops_module() -> ModuleType | None:
  function collect_autotune_configs (line 68) | def collect_autotune_configs() -> list[dict[str, Any]]:

FILE: src/axolotl/integrations/kernels/constants.py
  function resolve_moe_block_classes (line 47) | def resolve_moe_block_classes(model_type: str):

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/kernels/lora_ops.py
  function _next_power_of_2 (line 46) | def _next_power_of_2(n: int) -> int:
  function _block_r_for_rank (line 61) | def _block_r_for_rank(r: int) -> int:
  function round_expert_counts (line 71) | def round_expert_counts(
  function _get_smem_capacity (line 169) | def _get_smem_capacity() -> int:
  function _estimate_smem_usage (line 180) | def _estimate_smem_usage(
  function _estimate_register_pressure (line 198) | def _estimate_register_pressure(
  function _compute_expert_block_lora (line 234) | def _compute_expert_block_lora(
  function _scatter2scatter_lora_configs (line 356) | def _scatter2scatter_lora_configs():
  function _prune_fwd_configs (line 389) | def _prune_fwd_configs(configs, named_args, **kwargs):
  function _scatter2scatter_lora (line 463) | def _scatter2scatter_lora(
  function _scatter2scatter_lora_split (line 590) | def _scatter2scatter_lora_split(
  function scatter2scatter_lora (line 673) | def scatter2scatter_lora(
  function _compute_expert_block_lora_dX (line 807) | def _compute_expert_block_lora_dX(
  function _scatter2scatter_lora_dX_configs (line 936) | def _scatter2scatter_lora_dX_configs():
  function _prune_dX_configs (line 969) | def _prune_dX_configs(configs, named_args, **kwargs):
  function _scatter2scatter_lora_dX (line 1043) | def _scatter2scatter_lora_dX(
  function scatter2scatter_lora_dX (line 1166) | def scatter2scatter_lora_dX(
  function _group_bwd_lora_configs (line 1274) | def _group_bwd_lora_configs():
  function _prune_bwd_lora_configs (line 1308) | def _prune_bwd_lora_configs(configs, named_args, **kwargs):
  function _group_bwd_lora (line 1377) | def _group_bwd_lora(
  function _group_bwd_split_configs (line 1546) | def _group_bwd_split_configs():
  function _prune_split_configs (line 1565) | def _prune_split_configs(configs, named_args, **kwargs):
  function _group_bwd_lora_split (line 1617) | def _group_bwd_lora_split(
  function group_bwd_lora (line 1819) | def group_bwd_lora(
  function _group_bwd_lora_fused (line 1939) | def _group_bwd_lora_fused(
  function group_bwd_lora_fused (line 2145) | def group_bwd_lora_fused(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/kernels/ops.py
  function _compute_expert_block (line 18) | def _compute_expert_block(
  function _scatter2scatter_configs (line 62) | def _scatter2scatter_configs():
  function _scatter2scatter (line 79) | def _scatter2scatter(
  function scatter2scatter (line 169) | def scatter2scatter(
  function scatter2scatter_compileable (line 206) | def scatter2scatter_compileable(
  function _config_XtY (line 264) | def _config_XtY():
  function group_bwd_W (line 272) | def group_bwd_W(DY, X, expert_offsets, E, has_bias=False):
  function groupXtY_compileable (line 284) | def groupXtY_compileable(
  function _groupXtY (line 346) | def _groupXtY(
  function _xty_and_bias (line 469) | def _xty_and_bias(
  function _config_grouping (line 537) | def _config_grouping():
  function group (line 545) | def group(A, sorted_expert_idxs, coeff=None, fan_out=1, out=None):
  function group_compileable (line 558) | def group_compileable(
  function _group (line 595) | def _group(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/kernels/single.py
  function _single2scatter (line 13) | def _single2scatter(
  function single2scatter (line 70) | def single2scatter(X, W, expert_idxs):

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/layers.py
  function peft_lora_B_to_scattermoe (line 46) | def peft_lora_B_to_scattermoe(peft_B, num_experts, rank):
  function peft_lora_to_scattermoe (line 62) | def peft_lora_to_scattermoe(peft_A, peft_B, num_experts, rank):
  function peft_down_proj_lora_to_scattermoe (line 109) | def peft_down_proj_lora_to_scattermoe(peft_A, peft_B, num_experts, rank):
  function _unwrap_gate_lora (line 119) | def _unwrap_gate_lora(gate_module):
  function _convert_smoe_lora (line 155) | def _convert_smoe_lora(lora_A, lora_B, num_experts, rank, scaling):
  function _unwrap_experts_lora (line 161) | def _unwrap_experts_lora(experts_module):
  function _softmax_topk_route (line 228) | def _softmax_topk_route(
  function _sigmoid_topk_route (line 251) | def _sigmoid_topk_route(
  function _route (line 319) | def _route(moe_block, base_gate, hidden_states, gate_weight, gate_lora_d...
  function _compute_shared_expert (line 343) | def _compute_shared_expert(moe_block, hidden_states_flat):
  class ScatterMoEGatedMLP (line 380) | class ScatterMoEGatedMLP(nn.Module):
    method forward (line 381) | def forward(self, layer_input):
  class HFScatterMoEGatedMLP (line 434) | class HFScatterMoEGatedMLP(nn.Module):
    method forward (line 451) | def forward(self: nn.Module, layer_input: torch.Tensor):

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/lora_ops.py
  class ParallelExperts (line 20) | class ParallelExperts(nn.Module):
    method __init__ (line 29) | def __init__(
    method reset_parameters (line 50) | def reset_parameters(self) -> None:
    method extra_repr (line 55) | def extra_repr(self) -> str:
    method set_lora (line 62) | def set_lora(self, lora_A: torch.Tensor, lora_B: torch.Tensor, scaling...
    method clear_lora (line 68) | def clear_lora(self):
    method forward (line 74) | def forward(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/parallel_experts.py
  function compileable_bincount (line 16) | def compileable_bincount(x: torch.Tensor, minlength: int) -> torch.Tensor:
  function _ (line 21) | def _(x: torch.Tensor, minlength: int) -> torch.Tensor:
  function flatten_sort_count (line 26) | def flatten_sort_count(expert_idxs: torch.Tensor, num_experts: int):
  class ParallelLinear (line 37) | class ParallelLinear(torch.autograd.Function):
    method forward (line 39) | def forward(
    method backward (line 87) | def backward(ctx, grad_out: torch.Tensor):
  function parallel_linear (line 178) | def parallel_linear(
  class ParallelExperts (line 205) | class ParallelExperts(nn.Module):
    method __init__ (line 206) | def __init__(self, num_experts, input_size, output_size, bias=False) -...
    method extra_repr (line 220) | def extra_repr(self):
    method reset_parameters (line 225) | def reset_parameters(self) -> None:
    method forward (line 230) | def forward(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/parallel_linear_lora.py
  class ScatterMoELoRA (line 39) | class ScatterMoELoRA(torch.autograd.Function):
    method forward (line 50) | def forward(
    method backward (line 121) | def backward(ctx, grad_out: torch.Tensor):
  function _compute_lora_input_grad (line 342) | def _compute_lora_input_grad(
  function get_lora_params_from_wrapper (line 390) | def get_lora_params_from_wrapper(module) -> tuple:
  function parallel_linear_lora (line 425) | def parallel_linear_lora(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/selective_dequant.py
  function get_active_experts (line 28) | def get_active_experts(sorted_expert_idxs: torch.Tensor, E: int) -> torc...
  function remap_expert_indices (line 41) | def remap_expert_indices(
  function _selective_dequant_bnb4 (line 76) | def _selective_dequant_bnb4(
  function _selective_index_dense (line 178) | def _selective_index_dense(
  function selective_expert_weights (line 189) | def selective_expert_weights(
  function selective_lora_weights (line 255) | def selective_lora_weights(

FILE: src/axolotl/integrations/kernels/libs/scattermoe_lora/selective_dequant_kernel.py
  function _selective_dequant_nf4_kernel (line 45) | def _selective_dequant_nf4_kernel(
  function selective_dequant_nf4_triton (line 118) | def selective_dequant_nf4_triton(

FILE: src/axolotl/integrations/kernels/plugin.py
  function _check_sonicmoe_gpu_compat (line 13) | def _check_sonicmoe_gpu_compat():
  class KernelsPlugin (line 59) | class KernelsPlugin(BasePlugin):
    method get_input_args (line 60) | def get_input_args(self):
    method pre_model_load (line 63) | def pre_model_load(self, cfg):
    method _register_kernels (line 95) | def _register_kernels(self):
    method add_callbacks_pre_trainer (line 122) | def add_callbacks_pre_trainer(self, cfg, model):
    method _kernelize_model (line 132) | def _kernelize_model(self, model_type: str):

FILE: src/axolotl/integrations/kernels/sonicmoe/patch.py
  function patch_sonicmoe (line 34) | def patch_sonicmoe(model_type: str, torch_compile: bool = False):
  function _try_compile_routing (line 55) | def _try_compile_routing(routing_fn):
  function _patch_forward (line 69) | def _patch_forward(moe_cls, routing_fn, activation, router_attr):
  function _make_general_forward (line 99) | def _make_general_forward(moe_cls, routing_fn, activation):
  function _make_fused_forward (line 152) | def _make_fused_forward(moe_cls, activation, router_attr):
  function _compute_shared_expert (line 199) | def _compute_shared_expert(moe_block, hidden_states_flat):

FILE: src/axolotl/integrations/kernels/sonicmoe/routing.py
  function get_model_moe_config (line 18) | def get_model_moe_config(model_type: str):
  function softmax_topk_routing (line 81) | def softmax_topk_routing(
  function softmax_group_topk_routing (line 132) | def softmax_group_topk_routing(
  function sigmoid_topk_routing (line 188) | def sigmoid_topk_routing(

FILE: src/axolotl/integrations/kernels/sonicmoe/weight_converter.py
  function interleave_gate_up (line 23) | def interleave_gate_up(tensor: torch.Tensor) -> torch.Tensor:
  function deinterleave_gate_up (line 28) | def deinterleave_gate_up(tensor: torch.Tensor) -> torch.Tensor:
  class ConcatenatedToInterleaved (line 33) | class ConcatenatedToInterleaved(ConversionOps):
    method __init__ (line 42) | def __init__(self, dim: int = 1):
    method convert (line 46) | def convert(
    method _get_target_pattern (line 63) | def _get_target_pattern(
    method reverse_op (line 79) | def reverse_op(self) -> ConversionOps:
  class InterleavedToConcatenated (line 83) | class InterleavedToConcatenated(ConversionOps):
    method __init__ (line 92) | def __init__(self, dim: int = 1):
    method convert (line 96) | def convert(
    method _get_target_pattern (line 113) | def _get_target_pattern(
    method reverse_op (line 128) | def reverse_op(self) -> ConversionOps:
  function register_sonicmoe_weight_converter (line 132) | def register_sonicmoe_weight_converter(model_type: str):

FILE: src/axolotl/integrations/liger/args.py
  class LigerArgs (line 26) | class LigerArgs(BaseModel):
    method check_deprecated_swiglu (line 50) | def check_deprecated_swiglu(cls, data):
    method check_tiled_mlp_conflict (line 66) | def check_tiled_mlp_conflict(cls, data):
    method check_liger_rms_norm_tensor_parallel (line 79) | def check_liger_rms_norm_tensor_parallel(cls, data):
    method check_liger_use_token_scaling_flce (line 89) | def check_liger_use_token_scaling_flce(cls, data):
    method check_tensor_parallel_size_liger_fused_linear_cross_entropy (line 100) | def check_tensor_parallel_size_liger_fused_linear_cross_entropy(self):

FILE: src/axolotl/integrations/liger/models/base.py
  function lce_forward (line 18) | def lce_forward(
  function lce_maybe_trainable_lm_head (line 121) | def lce_maybe_trainable_lm_head(
  function _liger_for_causal_lm_loss (line 159) | def _liger_for_causal_lm_loss(
  function patch_lce_forward (line 172) | def patch_lce_forward(

FILE: src/axolotl/integrations/liger/models/deepseekv2.py
  function lce_forward (line 15) | def lce_forward(

FILE: src/axolotl/integrations/liger/models/jamba.py
  function lce_forward (line 19) | def lce_forward(

FILE: src/axolotl/integrations/liger/models/llama4.py
  function lce_forward (line 14) | def lce_forward(
  function apply_liger_kernel_to_llama4 (line 122) | def apply_liger_kernel_to_llama4(

FILE: src/axolotl/integrations/liger/models/qwen3.py
  function lce_forward (line 14) | def lce_forward(
  function apply_liger_kernel_to_qwen3 (line 109) | def apply_liger_kernel_to_qwen3(

FILE: src/axolotl/integrations/liger/models/qwen3_moe.py
  function lce_forward (line 15) | def lce_forward(
  function apply_liger_kernel_to_qwen3_moe (line 131) | def apply_liger_kernel_to_qwen3_moe(

FILE: src/axolotl/integrations/liger/plugin.py
  class LigerPlugin (line 14) | class LigerPlugin(BasePlugin):
    method get_input_args (line 19) | def get_input_args(self):
    method pre_model_load (line 22) | def pre_model_load(self, cfg):

FILE: src/axolotl/integrations/liger/utils.py
  function patch_with_compile_disable (line 10) | def patch_with_compile_disable(module, function_name):

FILE: src/axolotl/integrations/llm_compressor/args.py
  class CompressionArgs (line 11) | class CompressionArgs(BaseModel):
  class LLMCompressorArgs (line 32) | class LLMCompressorArgs(BaseModel):

FILE: src/axolotl/integrations/llm_compressor/plugin.py
  class LLMCompressorCallbackHandler (line 26) | class LLMCompressorCallbackHandler(TrainerCallback):
    method __init__ (line 33) | def __init__(self, trainer: Trainer, recipe: Any):
    method on_train_begin (line 50) | def on_train_begin(
    method on_step_begin (line 75) | def on_step_begin(
    method on_step_end (line 88) | def on_step_end(
    method on_train_end (line 103) | def on_train_end(
  class LLMCompressorPlugin (line 118) | class LLMCompressorPlugin(BasePlugin):
    method get_input_args (line 123) | def get_input_args(self) -> str:
    method add_callbacks_post_trainer (line 132) | def add_callbacks_post_trainer(self, cfg: Any, trainer: Trainer) -> list:
  function compute_loss_wrapper (line 151) | def compute_loss_wrapper(

FILE: src/axolotl/integrations/llm_compressor/utils.py
  function save_compressed_model (line 11) | def save_compressed_model(

FILE: src/axolotl/integrations/lm_eval/__init__.py
  class LMEvalPlugin (line 13) | class LMEvalPlugin(BasePlugin):
    method get_input_args (line 18) | def get_input_args(self):
    method post_train_unload (line 21) | def post_train_unload(self, cfg):

FILE: src/axolotl/integrations/lm_eval/args.py
  class LMEvalArgs (line 10) | class LMEvalArgs(BaseModel):

FILE: src/axolotl/integrations/lm_eval/cli.py
  function get_model_path (line 16) | def get_model_path(cfg: DictDefault) -> str | None:
  function build_lm_eval_command (line 31) | def build_lm_eval_command(
  function lm_eval (line 104) | def lm_eval(config: str, cloud: Optional[str] = None):

FILE: src/axolotl/integrations/spectrum/__init__.py
  function _generate_unfrozen_params_yaml (line 31) | def _generate_unfrozen_params_yaml(snr_data, top_fraction=0.5):
  class SpectrumPlugin (line 55) | class SpectrumPlugin(BasePlugin):
    method get_input_args (line 64) | def get_input_args(self):
    method pre_model_load (line 67) | def pre_model_load(self, cfg):

FILE: src/axolotl/integrations/spectrum/args.py
  class SpectrumArgs (line 24) | class SpectrumArgs(BaseModel):
    method check_fsdp_use_orig_params (line 34) | def check_fsdp_use_orig_params(cls, data):

FILE: src/axolotl/integrations/swanlab/args.py
  class SwanLabConfig (line 6) | class SwanLabConfig(BaseModel):
    method validate_swanlab_mode (line 96) | def validate_swanlab_mode(cls, v):
    method validate_swanlab_project (line 116) | def validate_swanlab_project(cls, v):
    method validate_swanlab_enabled_requires_project (line 128) | def validate_swanlab_enabled_requires_project(self):

FILE: src/axolotl/integrations/swanlab/callbacks.py
  class SwanLabRLHFCompletionCallback (line 20) | class SwanLabRLHFCompletionCallback(TrainerCallback):
    method __init__ (line 41) | def __init__(
    method on_init_end (line 61) | def on_init_end(
    method on_log (line 88) | def on_log(
    method on_train_end (line 166) | def on_train_end(

FILE: src/axolotl/integrations/swanlab/completion_logger.py
  class CompletionLogger (line 16) | class CompletionLogger:
    method __init__ (line 39) | def __init__(self, maxlen: int = 128):
    method add_dpo_completion (line 50) | def add_dpo_completion(
    method add_kto_completion (line 78) | def add_kto_completion(
    method add_orpo_completion (line 106) | def add_orpo_completion(
    method add_grpo_completion (line 134) | def add_grpo_completion(
    method log_to_swanlab (line 163) | def log_to_swanlab(self, table_name: str = "completions") -> bool:
    method clear (line 215) | def clear(self) -> None:
    method __len__ (line 219) | def __len__(self) -> int:
    method __repr__ (line 223) | def __repr__(self) -> str:

FILE: src/axolotl/integrations/swanlab/plugins.py
  class SwanLabPlugin (line 18) | class SwanLabPlugin(BasePlugin):
    method __init__ (line 35) | def __init__(self):
    method get_input_args (line 40) | def get_input_args(self) -> str:
    method register (line 44) | def register(self, cfg: dict):
    method pre_model_load (line 147) | def pre_model_load(self, cfg: DictDefault):
    method add_callbacks_pre_trainer (line 225) | def add_callbacks_pre_trainer(self, cfg: DictDefault, model):
    method post_trainer_create (line 261) | def post_trainer_create(self, cfg: DictDefault, trainer):
    method _get_swanlab_init_kwargs (line 298) | def _get_swanlab_init_kwargs(self, cfg: DictDefault) -> dict:
    method _prepare_config_for_logging (line 350) | def _prepare_config_for_logging(self, cfg: DictDefault) -> dict:
    method _register_lark_callback (line 435) | def _register_lark_callback(self, cfg: DictDefault):
    method _register_completion_callback (line 491) | def _register_completion_callback(self, cfg: DictDefault, trainer):

FILE: src/axolotl/integrations/swanlab/profiling.py
  function swanlab_profiling_context (line 18) | def swanlab_profiling_context(trainer: Any, func_name: str):
  function swanlab_profile (line 61) | def swanlab_profile(func: Callable) -> Callable:
  class ProfilingConfig (line 88) | class ProfilingConfig:
    method __init__ (line 99) | def __init__(
    method should_log (line 117) | def should_log(self, func_name: str, duration_seconds: float) -> bool:
  function swanlab_profiling_context_advanced (line 152) | def swanlab_profiling_context_advanced(

FILE: src/axolotl/kernels/geglu.py
  function _geglu_fwd_kernel (line 14) | def _geglu_fwd_kernel(
  function geglu_forward (line 45) | def geglu_forward(gate: torch.Tensor, up: torch.Tensor) -> torch.Tensor:
  function _geglu_bwd_kernel (line 71) | def _geglu_bwd_kernel(
  function geglu_backward (line 125) | def geglu_backward(

FILE: src/axolotl/kernels/lora.py
  function get_lora_parameters (line 23) | def get_lora_parameters(
  function matmul_lora (line 84) | def matmul_lora(
  class LoRA_MLP (line 132) | class LoRA_MLP(torch.autograd.Function):
    method forward (line 137) | def forward(
    method backward (line 219) | def backward(
  function apply_lora_mlp_swiglu (line 393) | def apply_lora_mlp_swiglu(self, X: torch.Tensor, inplace: bool = True) -...
  function apply_lora_mlp_geglu (line 436) | def apply_lora_mlp_geglu(self, X: torch.Tensor, inplace: bool = True) ->...
  class LoRA_QKV (line 478) | class LoRA_QKV(torch.autograd.Function):
    method forward (line 488) | def forward(
    method backward (line 555) | def backward(
  function apply_lora_qkv (line 698) | def apply_lora_qkv(
  class LoRA_O (line 740) | class LoRA_O(torch.autograd.Function):
    method forward (line 745) | def forward(
    method backward (line 783) | def backward(
  function apply_lora_o (line 831) | def apply_lora_o(self, X: torch.Tensor) -> torch.Tensor:

FILE: src/axolotl/kernels/quantize.py
  function dequantize_fp8 (line 18) | def dequantize_fp8(
  function dequantize (line 59) | def dequantize(

FILE: src/axolotl/kernels/swiglu.py
  function _swiglu_fwd_kernel (line 15) | def _swiglu_fwd_kernel(
  function _swiglu_bwd_kernel (line 50) | def _swiglu_bwd_kernel(
  function swiglu_forward (line 102) | def swiglu_forward(gate: torch.Tensor, up: torch.Tensor) -> torch.Tensor:
  function swiglu_backward (line 130) | def swiglu_backward(

FILE: src/axolotl/loaders/adapter.py
  function setup_quantized_meta_for_peft (line 30) | def setup_quantized_meta_for_peft(model: torch.nn.Module):
  function setup_quantized_peft_meta_for_training (line 42) | def setup_quantized_peft_meta_for_training(model: torch.nn.Module):
  function find_all_linear_names (line 50) | def find_all_linear_names(model):
  function load_lora (line 70) | def load_lora(
  function load_adapter (line 194) | def load_adapter(
  function load_llama_adapter (line 214) | def load_llama_adapter(

FILE: src/axolotl/loaders/model.py
  class ModelLoader (line 67) | class ModelLoader:
    method __init__ (line 98) | def __init__(
    method has_flash_attn (line 147) | def has_flash_attn(self) -> bool:
    method is_fsdp_enabled (line 152) | def is_fsdp_enabled(self):
    method is_qlora_and_fsdp_enabled (line 157) | def is_qlora_and_fsdp_enabled(self):
    method load (line 162) | def load(self) -> tuple[PreTrainedModel | PeftModelForCausalLM, PeftCo...
    method _apply_pre_model_load_setup (line 196) | def _apply_pre_model_load_setup(self):
    method _apply_post_model_load_setup (line 224) | def _apply_post_model_load_setup(self):
    method _configure_experts_implementation (line 241) | def _configure_experts_implementation(self):
    method _apply_activation_checkpointing (line 245) | def _apply_activation_checkpointing(self):
    method _resize_token_embeddings (line 254) | def _resize_token_embeddings(self):
    method _adjust_model_config (line 277) | def _adjust_model_config(self):
    method _configure_embedding_dtypes (line 306) | def _configure_embedding_dtypes(self):
    method _configure_qat (line 368) | def _configure_qat(self):
    method _load_adapters (line 381) | def _load_adapters(self) -> PeftConfig | None:
    method _apply_post_lora_load_setup (line 404) | def _apply_post_lora_load_setup(self, skip_move_to_device: bool):
    method _set_parallel_config (line 437) | def _set_parallel_config(self):
    method _set_auto_model_loader (line 444) | def _set_auto_model_loader(self):
    method _set_device_map_config (line 456) | def _set_device_map_config(self):
    method _set_quantization_config (line 539) | def _set_quantization_config(self):
    method _set_attention_config (line 623) | def _set_attention_config(self):
    method _check_model_requirements (line 650) | def _check_model_requirements(self):
    method _configure_zero3_memory_efficient_loading (line 660) | def _configure_zero3_memory_efficient_loading(
    method _load_model_from_config (line 699) | def _load_model_from_config(self, model_loader_class=None) -> PreTrain...
    method _load_model_from_pretrained (line 716) | def _load_model_from_pretrained(self, model_loader_class=None) -> PreT...
    method _build_model (line 726) | def _build_model(self) -> bool:
    method _set_z3_leaf_modules (line 845) | def _set_z3_leaf_modules(self):
    method _prepare_model_for_quantization (line 860) | def _prepare_model_for_quantization(self):
    method _convert_embedding_modules_dtype (line 897) | def _convert_embedding_modules_dtype(

FILE: src/axolotl/loaders/patch_manager.py
  class PatchManager (line 27) | class PatchManager:
    method apply_pre_config_load_patches (line 31) | def apply_pre_config_load_patches(cfg: DictDefault):
    method apply_pre_tokenizer_load_patches (line 52) | def apply_pre_tokenizer_load_patches(cfg: DictDefault):
    method __init__ (line 72) | def __init__(
    method has_flash_attn (line 90) | def has_flash_attn(self) -> bool:
    method apply_pre_model_load_patches (line 94) | def apply_pre_model_load_patches(self):
    method apply_post_plugin_pre_model_load_patches (line 122) | def apply_post_plugin_pre_model_load_patches(self):
    method _apply_transformers_patches (line 127) | def _apply_transformers_patches(self):
    method apply_post_model_build_patches (line 143) | def apply_post_model_build_patches(self, model: PreTrainedModel):
    method apply_post_model_load_patches (line 147) | def apply_post_model_load_patches(self, model: PreTrainedModel):
    method _apply_flash_attention_patches (line 154) | def _apply_flash_attention_patches(self):
    method _apply_chunked_cross_entropy_patch (line 162) | def _apply_chunked_cross_entropy_patch(self):
    method _apply_fsdp_patches (line 171) | def _apply_fsdp_patches(self):
    method _apply_adapter_patches (line 210) | def _apply_adapter_patches(self):
    method _apply_flex_attention_patches (line 217) | def _apply_flex_attention_patches(self):
    method _apply_sageattn_patches (line 227) | def _apply_sageattn_patches(self):
    method _apply_flash_attn_4_patches (line 234) | def _apply_flash_attn_4_patches(self):
    method _apply_model_specific_patches (line 243) | def _apply_model_specific_patches(self):
    method _apply_fp8_patches (line 294) | def _apply_fp8_patches(self):
    method _apply_flash_attention_peft_patches (line 305) | def _apply_flash_attention_peft_patches(self):
    method _apply_gradient_checkpointing_patches (line 314) | def _apply_gradient_checkpointing_patches(self):
    method _apply_mistral_cross_entropy_patch (line 337) | def _apply_mistral_cross_entropy_patch(self):
    method _apply_self_attention_lora_patch (line 349) | def _apply_self_attention_lora_patch(self):
    method _apply_multipack_patches (line 367) | def _apply_multipack_patches(self):
    method _apply_fsdp2_bnb_patches (line 405) | def _apply_fsdp2_bnb_patches(self):
    method _deactivate_hf_async_load (line 425) | def _deactivate_hf_async_load(self):
    method _apply_moe_expert_quantization_patch (line 430) | def _apply_moe_expert_quantization_patch(self):
    method _finalize_moe_expert_quantization (line 448) | def _finalize_moe_expert_quantization(self, model: PreTrainedModel):
    method _apply_tiled_mlp (line 469) | def _apply_tiled_mlp(self, model_type: str):
    method _apply_voxtral_patches (line 481) | def _apply_voxtral_patches(self):
    method _patch_attention (line 490) | def _patch_attention(self):
    method _patch_loss_llama (line 516) | def _patch_loss_llama(self):
    method _patch_llama_flash_attention (line 546) | def _patch_llama_flash_attention(self):
    method _patch_llama_xformers_attention (line 565) | def _patch_llama_xformers_attention(self):
    method _patch_llama_derived_model (line 574) | def _patch_llama_derived_model(self):
    method _apply_llama_flash_attn_patches (line 590) | def _apply_llama_flash_attn_patches(self, model):
    method _apply_unsloth_patches (line 610) | def _apply_unsloth_patches(self, model):
    method _apply_lora_kernel_patch (line 627) | def _apply_lora_kernel_patch(self, model):
    method _apply_patch_deepspeed_zero3 (line 638) | def _apply_patch_deepspeed_zero3(self):
    method _apply_apertus_patches (line 652) | def _apply_apertus_patches(self):
    method _apply_trl_vllm_patches (line 661) | def _apply_trl_vllm_patches(self):
    method _apply_trl_trainer_utils_patches (line 672) | def _apply_trl_trainer_utils_patches(self):
    method _apply_scaling_softmax_patch (line 705) | def _apply_scaling_softmax_patch(self, model: PreTrainedModel):

FILE: src/axolotl/loaders/processor.py
  function load_processor (line 17) | def load_processor(cfg: DictDefault, tokenizer: PreTrainedTokenizerBase):

FILE: src/axolotl/loaders/tokenizer.py
  function modify_tokenizer_files (line 30) | def modify_tokenizer_files(
  function load_tokenizer (line 130) | def load_tokenizer(cfg: DictDefault) -> PreTrainedTokenizer:

FILE: src/axolotl/loaders/utils.py
  function get_module_class_from_name (line 17) | def get_module_class_from_name(
  function check_model_config (line 45) | def check_model_config(cfg: DictDefault, model_config: PretrainedConfig):
  function load_model_config (line 151) | def load_model_config(cfg: DictDefault) -> PretrainedConfig | addict.Dict:
  function ensure_dtype (line 210) | def ensure_dtype(model: PreTrainedModel, dtype: torch.dtype = torch.bflo...
  function get_linear_embedding_layers (line 231) | def get_linear_embedding_layers(model_type: str) -> list[str]:

FILE: src/axolotl/logging_config.py
  class AxolotlOrWarnErrorFilter (line 15) | class AxolotlOrWarnErrorFilter(logging.Filter):
    method __init__ (line 22) | def __init__(self, **kwargs):
    method filter (line 40) | def filter(self, record: LogRecord) -> bool:
  class AxolotlLogger (line 51) | class AxolotlLogger(Logger):
    method __init__ (line 54) | def __init__(self, name: str, level: int = logging.NOTSET):
  class ColorfulFormatter (line 60) | class ColorfulFormatter(Formatter):
    method format (line 71) | def format(self, record):
  function configure_logging (line 142) | def configure_logging():

FILE: src/axolotl/models/mamba/__init__.py
  function check_mamba_ssm_installed (line 8) | def check_mamba_ssm_installed():
  function fix_mamba_attn_for_loss (line 16) | def fix_mamba_attn_for_loss():

FILE: src/axolotl/models/mamba/configuration_mamba.py
  class MambaConfig (line 8) | class MambaConfig(PretrainedConfig):
    method __init__ (line 15) | def __init__(

FILE: src/axolotl/models/mamba/modeling_mamba.py
  class MambaLMHeadModel (line 16) | class MambaLMHeadModel(nn.Module, GenerationMixin):
    method __init__ (line 17) | def __init__(
    method tie_weights (line 60) | def tie_weights(self):
    method allocate_inference_cache (line 63) | def allocate_inference_cache(self, batch_size, max_seqlen, dtype=None,...
    method forward (line 68) | def forward(
    method save_pretrained (line 110) | def save_pretrained(
    method from_pretrained (line 121) | def from_pretrained(cls, pretrained_model_name, device=None, dtype=Non...

FILE: src/axolotl/monkeypatch/accelerate/fsdp2.py
  function fsdp2_load_full_state_dict (line 20) | def fsdp2_load_full_state_dict(
  function get_state_dict (line 93) | def get_state_dict(self, model, unwrap=True):
  function patch_peft_param_wrapper_for_fsdp2 (line 189) | def patch_peft_param_wrapper_for_fsdp2():
  function _process_lora_module_for_fsdp (line 228) | def _process_lora_module_for_fsdp(module, fsdp2_kwargs):
  function fsdp2_prepare_model (line 272) | def fsdp2_prepare_model(accelerator, model: torch.nn.Module) -> torch.nn...
  function patch_tied_keys_for_meta_device (line 445) | def patch_tied_keys_for_meta_device():
  function patch_initialize_missing_keys_for_fsdp (line 482) | def patch_initialize_missing_keys_for_fsdp():
  function patch_accelerate_fsdp2 (line 522) | def patch_accelerate_fsdp2():

FILE: src/axolotl/monkeypatch/accelerate/parallelism_config.py
  function _validate_accelerator (line 11) | def _validate_accelerator(self, accelerator):
  function patched_is_fsdp2 (line 61) | def patched_is_fsdp2(self) -> bool:
  function patch_parallelism_config (line 73) | def patch_parallelism_config():
  function patch_prepare_cp (line 80) | def patch_prepare_cp():

FILE: src/axolotl/monkeypatch/attention/__init__.py
  function patch_xformers_attn_over_fa2 (line 8) | def patch_xformers_attn_over_fa2():
  function unpatch_xformers_attn_over_fa2 (line 16) | def unpatch_xformers_attn_over_fa2():

FILE: src/axolotl/monkeypatch/attention/flash_attn_4.py
  function _get_head_dims (line 10) | def _get_head_dims(model_config):
  function patch_flash_attn_4 (line 36) | def patch_flash_attn_4(model_config=None):

FILE: src/axolotl/monkeypatch/attention/flex_attn.py
  function patch_flex_wrapper (line 15) | def patch_flex_wrapper(**flex_attn_compile_kwargs):

FILE: src/axolotl/monkeypatch/attention/sage_attn.py
  function _is_sageattn_available (line 18) | def _is_sageattn_available():
  function _check_sageattn_imported (line 33) | def _check_sageattn_imported():
  function sage_attention_forward (line 42) | def sage_attention_forward(
  function patch_sageattn (line 193) | def patch_sageattn():

FILE: src/axolotl/monkeypatch/attention/xformers.py
  function xformers_attention_forward (line 19) | def xformers_attention_forward(

FILE: src/axolotl/monkeypatch/btlm_attn_hijack_flash.py
  function replace_btlm_attn_with_flash_attn (line 18) | def replace_btlm_attn_with_flash_attn(model_name="cerebras/btlm-3b-8k-ba...
  function flashattn_attn (line 31) | def flashattn_attn(

FILE: src/axolotl/monkeypatch/data/batch_dataset_fetcher.py
  class _MapDatasetFetcher (line 12) | class _MapDatasetFetcher(_BaseDatasetFetcher):
    method fetch (line 18) | def fetch(self, possibly_batched_index):
  function patch_fetchers (line 45) | def patch_fetchers():
  function patched_worker_loop (line 51) | def patched_worker_loop(*args, **kwargs):
  function apply_multipack_dataloader_patch (line 57) | def apply_multipack_dataloader_patch():
  function remove_multipack_dataloader_patch (line 79) | def remove_multipack_dataloader_patch():

FILE: src/axolotl/monkeypatch/deepspeed_utils.py
  function patch_checkpoint_wrapper_setattr (line 9) | def patch_checkpoint_wrapper_setattr():
  function apply_deepspeed_patches (line 60) | def apply_deepspeed_patches():

FILE: src/axolotl/monkeypatch/fsdp2_qlora.py
  function apply_init_sharded_param_patch (line 19) | def apply_init_sharded_param_patch():
  function apply_init_unsharded_param_patch (line 96) | def apply_init_unsharded_param_patch():
  function apply_linear8bitlt_save_patch (line 172) | def apply_linear8bitlt_save_patch():
  function apply_init_dtype_attrs_patch (line 205) | def apply_init_dtype_attrs_patch():

FILE: src/axolotl/monkeypatch/gradient_checkpointing/__init__.py
  function uses_gc_layers (line 19) | def uses_gc_layers(decoder_layer):
  function uses_gc_layers (line 24) | def uses_gc_layers(_):
  function hf_grad_checkpoint_offload_wrapper (line 28) | def hf_grad_checkpoint_offload_wrapper(decoder_layer, *args, use_reentra...
  function hf_grad_checkpoint_disk_offload_wrapper (line 45) | def hf_grad_checkpoint_disk_offload_wrapper(decoder_layer, *args, use_re...

FILE: src/axolotl/monkeypatch/gradient_checkpointing/offload_cpu.py
  class CPU_Offloaded_Gradient_Checkpointer (line 38) | class CPU_Offloaded_Gradient_Checkpointer(torch.autograd.Function):
    method forward (line 46) | def forward(ctx, forward_function, hidden_states, *args):
    method backward (line 57) | def backward(ctx, dY):

FILE: src/axolotl/monkeypatch/gradient_checkpointing/offload_disk.py
  class DiskOffloadManager (line 43) | class DiskOffloadManager:
    method __init__ (line 49) | def __init__(
    method _save_worker (line 106) | def _save_worker(self):
    method _save_tensor_to_disk (line 129) | def _save_tensor_to_disk(self, tensor: torch.Tensor, file_path: str):
    method _prefetch_worker (line 155) | def _prefetch_worker(self):
    method save_tensor (line 223) | def save_tensor(self, tensor: torch.Tensor):
    method wait_for_save (line 245) | def wait_for_save(self, file_path, timeout=None) -> None:
    method load_tensor (line 267) | def load_tensor(self, file_path, target_device="cuda"):
    method _safe_delete_file (line 308) | def _safe_delete_file(self, file_path):
    method trigger_prefetch (line 335) | def trigger_prefetch(self, n=None):
    method cleanup_tensor (line 356) | def cleanup_tensor(self, file_path: str):
    method cleanup (line 376) | def cleanup(self):
  class Disco (line 425) | class Disco(torch.autograd.Function):
    method get_instance (line 435) | def get_instance(prefetch_size=1, prefetch_to_gpu=True, save_workers=4):
    method forward (line 447) | def forward(
    method backward (line 481) | def backward(ctx, *grad_outputs):

FILE: src/axolotl/monkeypatch/llama_attn_hijack_flash.py
  function is_xformers_available (line 35) | def is_xformers_available() -> bool:
  function is_xformers_swiglu_available (line 39) | def is_xformers_swiglu_available() -> bool:
  function replace_llama_mlp_with_swiglu (line 54) | def replace_llama_mlp_with_swiglu(model):
  function patch_fa_llama_cross_entropy (line 68) | def patch_fa_llama_cross_entropy():
  function patch_llama_rms_norm (line 96) | def patch_llama_rms_norm():
  function replace_llama_attn_with_flash_attn (line 114) | def replace_llama_attn_with_flash_attn(
  function _prepare_decoder_attention_mask (line 136) | def _prepare_decoder_attention_mask(
  function flashattn_forward_with_s2attn (line 150) | def flashattn_forward_with_s2attn(

FILE: src/axolotl/monkeypatch/llama_attn_hijack_xformers.py
  function hijack_llama_attention (line 23) | def hijack_llama_attention():
  function xformers_forward (line 27) | def xformers_forward(

FILE: src/axolotl/monkeypatch/lora_kernels.py
  function original_apply_qkv (line 94) | def original_apply_qkv(
  function original_apply_o (line 115) | def original_apply_o(self: nn.Module, hidden_states: torch.Tensor) -> to...
  function get_attention_cls_from_config (line 131) | def get_attention_cls_from_config(cfg: DictDefault) -> Type[nn.Module]:
  function patch_self_attn_lora (line 201) | def patch_self_attn_lora(cfg: DictDefault):
  function find_self_attn_in_layer (line 265) | def find_self_attn_in_layer(
  function find_mlp_in_layer (line 277) | def find_mlp_in_layer(
  function get_layers (line 309) | def get_layers(model: PeftModelForCausalLM) -> list[nn.Module]:
  function apply_lora_kernel_patches (line 334) | def apply_lora_kernel_patches(
  class FakeMLP (line 476) | class FakeMLP(nn.Module):
    method __init__ (line 485) | def __init__(self, gate_proj, up_proj, down_proj):

FILE: src/axolotl/monkeypatch/loss/chunked.py
  class CEWithChunkedOutputLoss (line 12) | class CEWithChunkedOutputLoss(torch.nn.Module):
    method __init__ (line 19) | def __init__(self, num_output_chunks: int = 8, ignore_index: int = -100):
    method compute_cross_entropy (line 24) | def compute_cross_entropy(
    method forward (line 37) | def forward(
  function _build_chunked_ce_loss_fn (line 74) | def _build_chunked_ce_loss_fn(num_output_chunks: int = 8, ignore_index: ...
  function get_causal_lm_loss (line 82) | def get_causal_lm_loss(num_output_chunks: int = 8, ignore_index: int = -...
  function patch_chunked_ce_loss_fn (line 127) | def patch_chunked_ce_loss_fn(num_output_chunks: int = 8, ignore_index: i...

FILE: src/axolotl/monkeypatch/loss/eaft.py
  function eaft_loss (line 12) | def eaft_loss(outputs, labels, num_items_in_batch=None, alpha=1.0, k=20):

FILE: src/axolotl/monkeypatch/mistral_attn_hijack_flash.py
  function patch_mistral_cross_entropy (line 12) | def patch_mistral_cross_entropy():

FILE: src/axolotl/monkeypatch/mixtral/__init__.py
  function patch_mixtral_moe_forward_zero3 (line 8) | def patch_mixtral_moe_forward_zero3() -> None:

FILE: src/axolotl/monkeypatch/models/apertus/activation.py
  function patch_apertus_xielu_activation (line 6) | def patch_apertus_xielu_activation():

FILE: src/axolotl/monkeypatch/models/kimi_linear/configuration_kimi.py
  class KimiLinearConfig (line 13) | class KimiLinearConfig(PretrainedConfig):
    method __init__ (line 17) | def __init__(
    method is_mla (line 119) | def is_mla(self):
    method is_moe (line 130) | def is_moe(self):
    method is_linear_attn (line 134) | def is_linear_attn(self) -> bool:
    method is_kda_layer (line 144) | def is_kda_layer(self, layer_idx: int):

FILE: src/axolotl/monkeypatch/models/kimi_linear/modeling_kimi.py
  function load_balancing_loss_func (line 57) | def load_balancing_loss_func(
  class KimiDynamicCache (line 83) | class KimiDynamicCache:
    method __init__ (line 91) | def __init__(self, config: KimiLinearConfig):
    method __len__ (line 123) | def __len__(self):
    method update (line 126) | def update(
    method reorder_cache (line 146) | def reorder_cache(self, beam_idx: torch.LongTensor):
    method get_seq_length (line 172) | def get_seq_length(self, layer_idx: Optional[int] = 0) -> int:
    method get_mask_sizes (line 184) | def get_mask_sizes(
    method has_previous_state (line 199) | def has_previous_state(self):
  class KimiRMSNorm (line 206) | class KimiRMSNorm(nn.Module):
    method __init__ (line 207) | def __init__(self, hidden_size, eps=1e-6):
    method forward (line 215) | def forward(self, hidden_states):
  class KimiBlockSparseMLP (line 226) | class KimiBlockSparseMLP(nn.Module):
    method __init__ (line 227) | def __init__(
    method forward (line 243) | def forward(self, hidden_states):
  class KimiMLP (line 251) | class KimiMLP(nn.Module):
    method __init__ (line 252) | def __init__(
    method forward (line 266) | def forward(self, x):
  function repeat_kv (line 271) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function eager_attention_forward (line 285) | def eager_attention_forward(
  class KimiMLAAttention (line 315) | class KimiMLAAttention(nn.Module):
    method __init__ (line 320) | def __init__(self, config: KimiLinearConfig, layer_idx: int):
    method forward (line 372) | def forward(
  class KimiDeltaAttention (line 451) | class KimiDeltaAttention(nn.Module):
    method __init__ (line 452) | def __init__(self, config: KimiLinearConfig, layer_idx: int):
    method forward (line 510) | def forward(
  class KimiMoEGate (line 619) | class KimiMoEGate(nn.Module):
    method __init__ (line 625) | def __init__(self, config: KimiLinearConfig):
    method reset_parameters (line 635) | def reset_parameters(self) -> None:
    method forward (line 640) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  class KimiSparseMoeBlock (line 797) | class KimiSparseMoeBlock(nn.Module):
    method __init__ (line 803) | def __init__(self, config: KimiLinearConfig):
    method route_tokens_to_experts (line 830) | def route_tokens_to_experts(
    method forward (line 887) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
    method _training_forward (line 922) | def _training_forward(
    method _inference_forward (line 994) | def _inference_forward(
  class KimiDecoderLayer (line 1037) | class KimiDecoderLayer(nn.Module):
    method __init__ (line 1038) | def __init__(self, config: KimiLinearConfig, layer_idx: int):
    method forward (line 1063) | def forward(
  class KimiPreTrainedModel (line 1127) | class KimiPreTrainedModel(PreTrainedModel):
    method _init_weights (line 1141) | def _init_weights(self, module):
  class KimiLinearModel (line 1153) | class KimiLinearModel(KimiPreTrainedModel):
    method __init__ (line 1154) | def __init__(self, config: KimiLinearConfig):
    method _update_linear_attn_mask (line 1185) | def _update_linear_attn_mask(self, attention_mask, cache_position):
    method forward (line 1199) | def forward(
  class KimiLinearForCausalLM (line 1272) | class KimiLinearForCausalLM(KimiPreTrainedModel, GenerationMixin):
    method __init__ (line 1275) | def __init__(self, config):
    method forward (line 1285) | def forward(

FILE: src/axolotl/monkeypatch/models/kimi_linear/patch_kimi_linear.py
  function get_patch_file_path (line 13) | def get_patch_file_path(package_dot_path: str, filename: str) -> Path:
  function _load_local_module (line 23) | def _load_local_module(module_name: str, filename: str):
  function _patch_get_class_in_module (line 38) | def _patch_get_class_in_module():
  function patch_kimi (line 73) | def patch_kimi():

FILE: src/axolotl/monkeypatch/models/kimi_linear/tokenization_kimi.py
  class TikTokenTokenizer (line 33) | class TikTokenTokenizer(PreTrainedTokenizer):
    method __init__ (line 78) | def __init__(
    method encode (line 167) | def encode(
    method decode (line 232) | def decode(self, token_ids: Union[int, List[int]], **kwargs) -> str:
    method _split_whitespaces_or_nonwhitespaces (line 253) | def _split_whitespaces_or_nonwhitespaces(
    method pre_tokenizer_process (line 278) | def pre_tokenizer_process(self, text: str) -> List[str]:
    method vocab_size (line 288) | def vocab_size(self) -> int:
    method get_vocab (line 291) | def get_vocab(self) -> Dict[str, int]:
    method _tokenize (line 294) | def _tokenize(self, text: str, **kwargs) -> List[str]:
    method _convert_token_to_id (line 297) | def _convert_token_to_id(self, token: str) -> int:
    method _convert_id_to_token (line 300) | def _convert_id_to_token(self, index: int) -> str:
    method clean_up_tokenization (line 304) | def clean_up_tokenization(out_string: str) -> str:
    method convert_tokens_to_string (line 307) | def convert_tokens_to_string(self, tokens: List[str]) -> str:
    method save_vocabulary (line 314) | def save_vocabulary(
    method apply_chat_template (line 334) | def apply_chat_template(
  function deep_sort_dict (line 352) | def deep_sort_dict(obj: Any) -> Any:

FILE: src/axolotl/monkeypatch/models/llama4/modeling.py
  class Llama4TextExperts (line 13) | class Llama4TextExperts(nn.Module):
    method __init__ (line 18) | def __init__(self, config: Llama4Config):
    method forward (line 50) | def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
  function patch_llama4_linearized_modeling (line 90) | def patch_llama4_linearized_modeling():

FILE: src/axolotl/monkeypatch/models/mistral3/mistral_common_tokenizer.py
  function apply_mistral_tokenizer_image_patch (line 14) | def apply_mistral_tokenizer_image_patch():

FILE: src/axolotl/monkeypatch/models/pixtral/modeling_flash_attention_utils.py
  function apply_patch_is_packed_sequence (line 6) | def apply_patch_is_packed_sequence():

FILE: src/axolotl/monkeypatch/models/qwen3_5/modeling.py
  function get_cu_seqlens (line 24) | def get_cu_seqlens(position_ids):
  function _inject_fla_kernels (line 49) | def _inject_fla_kernels(module) -> None:
  function _patched_decoder_forward (line 68) | def _patched_decoder_forward(
  function _make_qwen3_5_gated_delta_forward (line 113) | def _make_qwen3_5_gated_delta_forward(apply_mask_fn):
  function _apply_packing_patches (line 240) | def _apply_packing_patches(model_type: str, cls_prefix: str, forward_fac...
  function patch_qwen3_5_modeling_packing (line 260) | def patch_qwen3_5_modeling_packing():
  function patch_qwen3_5_moe_modeling_packing (line 264) | def patch_qwen3_5_moe_modeling_packing():
  function patch_qwen3_5_vlm_flash_attention (line 270) | def patch_qwen3_5_vlm_flash_attention():

FILE: src/axolotl/monkeypatch/models/qwen3_next/modeling.py
  function get_cu_seqlens (line 18) | def get_cu_seqlens(position_ids):
  function patch_qwen3_next_decoder_layer (line 39) | def patch_qwen3_next_decoder_layer():
  function patch_qwen3_next_gateddelta_layer (line 110) | def patch_qwen3_next_gateddelta_layer():
  function patch_qwen3_next_imports (line 283) | def patch_qwen3_next_imports():
  function patch_qwen3_next_modeling_packing (line 336) | def patch_qwen3_next_modeling_packing():

FILE: src/axolotl/monkeypatch/models/voxtral/modeling.py
  function patch_voxtral_conditional_generation_forward (line 10) | def patch_voxtral_conditional_generation_forward():

FILE: src/axolotl/monkeypatch/moe_quant.py
  class Bnb8bitParametrization (line 23) | class Bnb8bitParametrization(torch.nn.Module):
    method __init__ (line 26) | def __init__(self, row_stats: torch.Tensor):
    method forward (line 31) | def forward(self, quantized_param: torch.Tensor) -> torch.Tensor:
  function _enable_parametrization_cache (line 40) | def _enable_parametrization_cache(module, inputs):
  function _disable_parametrization_cache (line 44) | def _disable_parametrization_cache(module, inputs, output):
  function replace_parameter_8bit (line 50) | def replace_parameter_8bit(module, param_name):
  function patch_moe_quantization_on_load (line 71) | def patch_moe_quantization_on_load(cfg):
  function get_moe_quantized_count (line 146) | def get_moe_quantized_count():
  function patch_peft_target_parameters_matching (line 151) | def patch_peft_target_parameters_matching():

FILE: src/axolotl/monkeypatch/multipack.py
  function patch_for_multipack (line 66) | def patch_for_multipack(model_type, model_name=None, has_remote_code=Fal...
  function patch_remote (line 80) | def patch_remote(model_name):

FILE: src/axolotl/monkeypatch/peft/utils.py
  function get_peft_prep_code (line 32) | def get_peft_prep_code() -> str:
  function check_peft_prep_code_is_patchable (line 37) | def check_peft_prep_code_is_patchable() -> bool:
  function patch_peft_prep_code (line 43) | def patch_peft_prep_code():

FILE: src/axolotl/monkeypatch/relora.py
  function magnitude_pruning_ (line 33) | def magnitude_pruning_(tensor, prune_ratio):
  function reset_optimizer (line 43) | def reset_optimizer(
  class ReLoRACallback (line 81) | class ReLoRACallback(TrainerCallback):
    method __init__ (line 84) | def __init__(self, cfg: DictDefault):
    method on_train_begin (line 101) | def on_train_begin(
    method on_step_begin (line 120) | def on_step_begin(
    method on_save (line 182) | def on_save(
    method on_log (line 220) | def on_log(
    method on_train_end (line 231) | def on_train_end(
  function sharded_paths (line 255) | def sharded_paths(path: str, module_names: List[str]) -> Dict[str, str]:
  function lora_delta_weight (line 270) | def lora_delta_weight(layer: peft.tuners.lora.LoraLayer, device) -> torc...
  function find_lora_modules (line 289) | def find_lora_modules(model: peft.LoraModel) -> Dict[str, peft.tuners.lo...
  function update_weights (line 305) | def update_weights(
  function merge_and_save (line 330) | def merge_and_save(
  function load_weight_checkpoint (line 420) | def load_weight_checkpoint(model: peft.LoraModel, checkpoint_path: str):

FILE: src/axolotl/monkeypatch/ring_attn/adapters/batch.py
  function create_flash_attn_forward_varlen_llama3 (line 42) | def create_flash_attn_forward_varlen_llama3(
  function substitute_hf_flash_attn (line 156) | def substitute_hf_flash_attn(

FILE: src/axolotl/monkeypatch/ring_attn/patch.py
  function get_ring_attn_group (line 37) | def get_ring_attn_group() -> dist.ProcessGroup:
  function set_ring_attn_group (line 44) | def set_ring_attn_group(ring_attn_group: dist.ProcessGroup | None):
  function create_ring_flash_attention_forward (line 50) | def create_ring_flash_attention_forward(
  function register_ring_attn_from_device_mesh (line 135) | def register_ring_attn_from_device_mesh(
  function update_ring_attn_params (line 214) | def update_ring_attn_params(position_ids: torch.Tensor | None):

FILE: src/axolotl/monkeypatch/scaled_softmax_attn.py
  function patch_scaled_softmax_attention (line 29) | def patch_scaled_softmax_attention(
  function ssmax_flex_attention_forward (line 53) | def ssmax_flex_attention_forward(
  function unpatch_scaled_softmax_attention (line 133) | def unpatch_scaled_softmax_attention():

FILE: src/axolotl/monkeypatch/stablelm_attn_hijack_flash.py
  function replace_stablelm_attn_with_flash_attn (line 42) | def replace_stablelm_attn_with_flash_attn(model_name="stabilityai/stable...
  function rotate_half (line 57) | def rotate_half(x: torch.Tensor):
  function apply_rotary_pos_emb (line 64) | def apply_rotary_pos_emb(q, k, cos, sin, position_ids):
  function repeat_kv (line 76) | def repeat_kv(hidden_states: torch.Tensor, n_rep: int) -> torch.Tensor:
  function flashattn_attn (line 90) | def flashattn_attn(
  function decoder_layer_forward (line 200) | def decoder_layer_forward(
  function stablelm_model_forward (line 247) | def stablelm_model_forward(

FILE: src/axolotl/monkeypatch/tiled_mlp/base.py
  class DeepSpeedTiledMLPMoE (line 11) | class DeepSpeedTiledMLPMoE(torch.autograd.Function):
    method forward (line 13) | def forward(
    method backward (line 47) | def backward(ctx, *grads) -> torch.Tensor:
  class TiledMLP (line 99) | class TiledMLP(torch.autograd.Function):
    method forward (line 105) | def forward(
    method backward (line 138) | def backward(ctx, *grads) -> torch.Tensor:
  class GradientAccumulator (line 191) | class GradientAccumulator:
    method __init__ (line 197) | def __init__(
    method install_hooks (line 223) | def install_hooks(self, is_last_shard: bool):
    method cleanup (line 251) | def cleanup(self):

FILE: src/axolotl/monkeypatch/tiled_mlp/patch.py
  function patch_tiled_mlp (line 15) | def patch_tiled_mlp(model_type, use_original_mlp=True, cfg_num_shards=No...

FILE: src/axolotl/monkeypatch/trainer/lr.py
  function _get_learning_rate (line 13) | def _get_learning_rate(self):
  function patch_trainer_get_lr (line 39) | def patch_trainer_get_lr():

FILE: src/axolotl/monkeypatch/trainer/trl.py
  function prepare_fsdp (line 4) | def prepare_fsdp(model, accelerator):
  function patch_trl_prepare_fsdp2 (line 10) | def patch_trl_prepare_fsdp2():

FILE: src/axolotl/monkeypatch/trainer/trl_vllm.py
  function _batch_update_named_params (line 20) | def _batch_update_named_params(
  function _update_model_params (line 65) | def _update_model_params(self, model: nn.Module, chunk_size: int | None ...
  function _patched_extract_logprobs (line 71) | def _patched_extract_logprobs(all_outputs):
  function _patched_split_tensor_dict (line 97) | def _patched_split_tensor_dict(tensor_dict, num_chunks):
  function _patched_shuffle_sequence_dict (line 121) | def _patched_shuffle_sequence_dict(seq_dict):
  function _patch_sync_weights_batched (line 146) | def _patch_sync_weights_batched(original_init):
  function _make_batched_sync_weights (line 157) | def _make_batched_sync_weights(original_sync_weights):
  function patch_trl_vllm (line 214) | def patch_trl_vllm():

FILE: src/axolotl/monkeypatch/trainer/utils.py
  function _entropy_online_kernel (line 22) | def _entropy_online_kernel(
  function _entropy_online_kernel_strided (line 62) | def _entropy_online_kernel_strided(
  function entropy_from_logits (line 108) | def entropy_from_logits(logits: torch.Tensor, chunk_size: int = 128) -> ...
  function selective_log_softmax_original (line 174) | def selective_log_softmax_original(logits, index) -> torch.Tensor:
  function _selective_logsoftmax_fwd_kernel (line 199) | def _selective_logsoftmax_fwd_kernel(
  function _selective_logsoftmax_bwd_kernel (line 255) | def _selective_logsoftmax_bwd_kernel(
  class _SelectiveLogSoftmaxTriton (line 320) | class _SelectiveLogSoftmaxTriton(torch.autograd.Function):
    method forward (line 322) | def forward(ctx, flat_logits, flat_index, K, K_BLOCK, V, BLOCK_V, MAX_...
    method backward (line 352) | def backward(ctx, grad_output):
  function selective_log_softmax (line 390) | def selective_log_softmax(logits, index) -> torch.Tensor:

FILE: src/axolotl/monkeypatch/trainer_accelerator_args.py
  function get_create_accelerate_code (line 30) | def get_create_accelerate_code() -> str:
  function check_create_accelerate_code_is_patchable (line 35) | def check_create_accelerate_code_is_patchable() -> bool:
  function patch_create_accelerate_code_for_fp8 (line 41) | def patch_create_accelerate_code_for_fp8(enable_fsdp_float8_all_gather: ...

FILE: src/axolotl/monkeypatch/trainer_fsdp_optim.py
  function get_training_loop_code (line 25) | def get_training_loop_code() -> str:
  function check_training_loop_is_patchable (line 30) | def check_training_loop_is_patchable() -> bool:
  function patch_training_loop_for_fsdp (line 36) | def patch_training_loop_for_fsdp():

FILE: src/axolotl/monkeypatch/transformers/trainer_context_parallel.py
  function patch_prepare_context_parallel_inputs (line 19) | def patch_prepare_context_parallel_inputs() -> None:

FILE: src/axolotl/monkeypatch/transformers/trainer_loss_calc.py
  function check_evaluation_loop_is_patchable (line 39) | def check_evaluation_loop_is_patchable() -> bool:
  function patch_evaluation_loop (line 44) | def patch_evaluation_loop():
  function check_maybe_log_save_evaluate_is_patchable (line 95) | def check_maybe_log_save_evaluate_is_patchable() -> bool:
  function patch_maybe_log_save_evaluate (line 100) | def patch_maybe_log_save_evaluate():

FILE: src/axolotl/monkeypatch/transformers_fa_utils.py
  function fixed_fa_peft_integration_check (line 15) | def fixed_fa_peft_integration_check(
  function patch_fa_peft_integration (line 63) | def patch_fa_peft_integration():

FILE: src/axolotl/monkeypatch/unsloth_.py
  function original_apply_qkv (line 35) | def original_apply_qkv(self, hidden_states):
  function original_apply_o (line 42) | def original_apply_o(self, hidden_states):
  function get_self_attn_code (line 47) | def get_self_attn_code() -> str:
  function check_self_attn_is_patchable (line 52) | def check_self_attn_is_patchable() -> bool:
  function integrate_cross_entropy_loss_patch (line 58) | def integrate_cross_entropy_loss_patch(model_type: str = "llama") -> None:
  function patch_self_attn_lora (line 91) | def patch_self_attn_lora():
  function integrate_rope_embeddings (line 130) | def integrate_rope_embeddings():
  function integrate_lora_mlp_patch (line 148) | def integrate_lora_mlp_patch(peft_model: PeftModelForCausalLM):
  function integrate_lora_patch (line 183) | def integrate_lora_patch(peft_model: PeftModelForCausalLM, cfg):
  function patch_unsloth_layernorm (line 228) | def patch_unsloth_layernorm():

FILE: src/axolotl/monkeypatch/utils.py
  function get_max_seqlen_in_batch (line 13) | def get_max_seqlen_in_batch(attention_mask: torch.Tensor) -> torch.Tensor:
  function get_unpad_data (line 26) | def get_unpad_data(attention_mask: torch.Tensor):
  function get_cu_seqlens (line 43) | def get_cu_seqlens(attn_mask):
  function get_cu_seqlens_from_pos_ids (line 94) | def get_cu_seqlens_from_pos_ids(
  function set_module_name (line 155) | def set_module_name(model, name, value):
  function detab_code (line 168) | def detab_code(code: str) -> Tuple[str, str]:

FILE: src/axolotl/monkeypatch/xformers_/__init__.py
  class FusedMLP (line 12) | class FusedMLP(torch.nn.Module):
    method __init__ (line 17) | def __init__(
    method _post_training (line 38) | def _post_training(self, model, name):
    method forward (line 51) | def forward(self, x: torch.Tensor) -> torch.Tensor:

FILE: src/axolotl/processing_strategies.py
  class ProcessingStrategy (line 21) | class ProcessingStrategy:
    method __init__ (line 24) | def __init__(
    method __call__ (line 47) | def __call__(self, examples: list[dict]) -> list[dict]:
    method _mask_non_assistant (line 223) | def _mask_non_assistant(self, labels: Tensor) -> Tensor:
    method process_labels (line 230) | def process_labels(self, input_ids: Tensor) -> Tensor:
  class Qwen2VLProcessingStrategy (line 244) | class Qwen2VLProcessingStrategy(ProcessingStrategy):
    method __init__ (line 247) | def __init__(
  class Qwen3_5ProcessingStrategy (line 261) | class Qwen3_5ProcessingStrategy(ProcessingStrategy):
    method __init__ (line 264) | def __init__(
    method process_labels (line 281) | def process_labels(self, input_ids):
  class Gemma3ProcessingStrategy (line 287) | class Gemma3ProcessingStrategy(ProcessingStrategy):
    method __init__ (line 290) | def __init__(
    method process_labels (line 303) | def process_labels(self, input_ids):
  class Gemma3nProcessingStrategy (line 314) | class Gemma3nProcessingStrategy(ProcessingStrategy):
    method _mask_non_assistant (line 317) | def _mask_non_assistant(self, labels: Tensor) -> Tensor:
    method process_labels (line 389) | def process_labels(self, input_ids):
  class VoxtralProcessingStrategy (line 407) | class VoxtralProcessingStrategy(ProcessingStrategy):
    method __init__ (line 410) | def __init__(
    method process_labels (line 425) | def process_labels(self, input_ids):
  class SmolVLM2ProcessingStrategy (line 435) | class SmolVLM2ProcessingStrategy(ProcessingStrategy):
    method __init__ (line 438) | def __init__(
  class Mistral3ProcessingStrategy (line 453) | class Mistral3ProcessingStrategy(ProcessingStrategy):
    method __init__ (line 456) | def __init__(
    method process_labels (line 472) | def process_labels(self, input_ids):
  class InternVLProcessingStrategy (line 483) | class InternVLProcessingStrategy(ProcessingStrategy):
    method __init__ (line 486) | def __init__(
    method process_labels (line 500) | def process_labels(self, input_ids):
  class Glm4vProcessingStrategy (line 514) | class Glm4vProcessingStrategy(ProcessingStrategy):
    method __init__ (line 517) | def __init__(
    method process_labels (line 550) | def process_labels(self, input_ids):
  function get_processing_strategy (line 566) | def get_processing_strategy(

FILE: src/axolotl/prompt_strategies/__init__.py
  function load (line 12) | def load(strategy, tokenizer, cfg, ds_cfg, processor=None):

FILE: src/axolotl/prompt_strategies/alpaca_chat.py
  function load (line 12) | def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):
  class AlpacaConcisePrompter (line 25) | class AlpacaConcisePrompter(AlpacaPrompter):
  class AlpacaChatPrompter (line 34) | class AlpacaChatPrompter(AlpacaPrompter):
    method __init__ (line 42) | def __init__(self):
  class NoSystemPrompter (line 47) | class NoSystemPrompter(AlpacaPrompter):
    method __init__ (line 57) | def __init__(self):
  class AlpacaQAPromptTokenizingStrategy (line 61) | class AlpacaQAPromptTokenizingStrategy(InstructionPromptTokenizingStrate...
    method parse_instruction_fields (line 66) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class CamelAIPromptTokenizingStrategy (line 74) | class CamelAIPromptTokenizingStrategy(InstructionPromptTokenizingStrategy):
    method parse_instruction_fields (line 79) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  function load_concise (line 87) | def load_concise(tokenizer, cfg):
  function load_qa (line 96) | def load_qa(tokenizer, cfg):
  function load_camel_ai (line 105) | def load_camel_ai(tokenizer, cfg):
  function load_no_prompt (line 114) | def load_no_prompt(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/alpaca_instruct.py
  function load (line 7) | def load(tokenizer, cfg):
  function load_no_prompt (line 16) | def load_no_prompt(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/alpaca_w_system.py
  class InstructionWSystemPromptTokenizingStrategy (line 11) | class InstructionWSystemPromptTokenizingStrategy(PromptTokenizingStrategy):
    method parse_instruction_fields (line 16) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str, str]:
    method tokenize_prompt (line 24) | def tokenize_prompt(self, prompt):
  class SystemDataPrompter (line 55) | class SystemDataPrompter(AlpacaPrompter):
    method build_prompt_w_system (line 62) | def build_prompt_w_system(
  class OpenOrcaSystemDataPrompter (line 89) | class OpenOrcaSystemDataPrompter(SystemDataPrompter):
    method match_prompt_style (line 94) | def match_prompt_style(self):
  class OpenOrcaPromptTokenizingStrategy (line 111) | class OpenOrcaPromptTokenizingStrategy(InstructionWSystemPromptTokenizin...
    method parse_instruction_fields (line 116) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str, str]:
  function load (line 125) | def load(tokenizer, cfg):
  function load_instruct (line 129) | def load_instruct(tokenizer, cfg):
  function load_chat (line 138) | def load_chat(tokenizer, cfg):
  function load_open_orca (line 147) | def load_open_orca(tokenizer, cfg):
  function load_open_orca_chatml (line 156) | def load_open_orca_chatml(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/base.py
  function load (line 12) | def load(strategy, cfg, module_base=None, **kwargs):

FILE: src/axolotl/prompt_strategies/bradley_terry/__init__.py
  function load (line 12) | def load(strategy, tokenizer, cfg, ds_cfg):

FILE: src/axolotl/prompt_strategies/bradley_terry/chat_template.py
  class BTChatTemplateStrategy (line 19) | class BTChatTemplateStrategy(ChatTemplateStrategy):
    method supports_batched (line 25) | def supports_batched(self) -> bool:
    method _tokenize_single_prompt (line 28) | def _tokenize_single_prompt(self, prompt):
  function load (line 83) | def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):

FILE: src/axolotl/prompt_strategies/bradley_terry/llama3.py
  function icr (line 6) | def icr(

FILE: src/axolotl/prompt_strategies/chat_template.py
  class ChatTemplatePrompter (line 28) | class ChatTemplatePrompter(Prompter):
    method __init__ (line 31) | def __init__(
    method chat_template_msg_variables (line 89) | def chat_template_msg_variables(self) -> Set[str]:
    method build_prompt (line 92) | def build_prompt(
    method get_offsets_for_train_detail (line 158) | def get_offsets_for_train_detail(
    method adjust_train_details (line 202) | def adjust_train_details(
    method get_chat_template_msg_variables (line 260) | def get_chat_template_msg_variables(
  class ChatTemplateStrategy (line 267) | class ChatTemplateStrategy(PromptTokenizingStrategy):
    method __init__ (line 272) | def __init__(
    method _validate_eot_and_eos_tokens (line 318) | def _validate_eot_and_eos_tokens(self):
    method supports_batched (line 384) | def supports_batched(self) -> bool:
    method is_prompt_batched (line 388) | def is_prompt_batched(self, prompt: dict[str, Any]) -> bool:
    method tokenize_prompt (line 396) | def tokenize_prompt(self, prompt: dict[str, Any]):
    method _tokenize_single_prompt (line 423) | def _tokenize_single_prompt(self, prompt: dict) -> Dict[str, List[int]]:
    method find_first_eos_token (line 587) | def find_first_eos_token(self, input_ids, start_idx):
    method find_first_eot_token (line 594) | def find_first_eot_token(self, input_ids, start_idx):
    method find_turn (line 613) | def find_turn(
    method get_conversation_thread (line 700) | def get_conversation_thread(self, prompt):
    method transform_message (line 733) | def transform_message(self, message: dict) -> dict:
    method _get_images (line 828) | def _get_images(self, prompt):
    method _get_tools (line 831) | def _get_tools(self, prompt) -> list[dict] | None:
    method _get_messages (line 862) | def _get_messages(self, prompt):
  class MistralStrategy (line 876) | class MistralStrategy(ChatTemplateStrategy):
    method __init__ (line 881) | def __init__(
    method find_first_eot_token (line 931) | def find_first_eot_token(self, input_ids, start_idx):
  class MistralPrompter (line 937) | class MistralPrompter(ChatTemplatePrompter):
    method __init__ (line 942) | def __init__(self, *args, **kwargs):
  class StrategyLoader (line 948) | class StrategyLoader:
    method _get_strategy_cls (line 953) | def _get_strategy_cls(self, cfg):
    method _get_prompter_cls (line 959) | def _get_prompter_cls(self, cfg):
    method _get_strategy_params (line 965) | def _get_strategy_params(self, cfg, ds_cfg: Dict[str, Any]):
    method __call__ (line 976) | def __call__(

FILE: src/axolotl/prompt_strategies/completion.py
  class CompletionPromptTokenizingStrategy (line 11) | class CompletionPromptTokenizingStrategy(InstructionPromptTokenizingStra...
    method __init__ (line 18) | def __init__(self, *args, max_length=None, **kwargs):
    method supports_batched (line 24) | def supports_batched(self):
    method field (line 28) | def field(self) -> str:
    method field (line 32) | def field(self, new_field: str):
    method parse_instruction_fields (line 35) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
    method tokenize_prompt (line 42) | def tokenize_prompt(self, prompt):
    method _build_full_prompt (line 62) | def _build_full_prompt(self, instruction, input, response):
  class CompletionPrompter (line 66) | class CompletionPrompter:
    method build_prompt (line 71) | def build_prompt(
  function load (line 80) | def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):

FILE: src/axolotl/prompt_strategies/context_qa.py
  function load_404 (line 10) | def load_404(tokenizer, cfg):
  function load (line 19) | def load(tokenizer, cfg):
  function load_v2 (line 28) | def load_v2(tokenizer, cfg):
  class AlpacaContextPrompter (line 37) | class AlpacaContextPrompter(AlpacaPrompter):
  class AlpacaContextPromptTokenizingStrategy (line 50) | class AlpacaContextPromptTokenizingStrategy(InstructionPromptTokenizingS...
    method parse_instruction_fields (line 55) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class ContextQaV2PromptTokenizingStrategy (line 63) | class ContextQaV2PromptTokenizingStrategy(InstructionPromptTokenizingStr...
    method parse_instruction_fields (line 68) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class ContextV2Prompter (line 80) | class ContextV2Prompter(AlpacaPrompter):
    method match_prompt_style (line 88) | def match_prompt_style(self):
  class AlpacaMissingInfoContextPromptTokenizingStrategy (line 94) | class AlpacaMissingInfoContextPromptTokenizingStrategy(
    method parse_instruction_fields (line 102) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:

FILE: src/axolotl/prompt_strategies/creative_acr.py
  class CreativeAnsweringPromptTokenizingStrategy (line 10) | class CreativeAnsweringPromptTokenizingStrategy(InstructionPromptTokeniz...
    method parse_instruction_fields (line 15) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class CreativeCritiquePromptTokenizingStrategy (line 27) | class CreativeCritiquePromptTokenizingStrategy(InstructionPromptTokenizi...
    method parse_instruction_fields (line 63) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class CreativeRevisePromptTokenizingStrategy (line 84) | class CreativeRevisePromptTokenizingStrategy(InstructionPromptTokenizing...
    method parse_instruction_fields (line 103) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class CreativePrompterBase (line 126) | class CreativePrompterBase:
    method build_prompt (line 134) | def build_prompt(
  class CreativeAnswerPrompter (line 149) | class CreativeAnswerPrompter(CreativePrompterBase):
  class CreativeCritiquePrompter (line 157) | class CreativeCritiquePrompter(CreativePrompterBase):
  class CreativeRevisePrompter (line 165) | class CreativeRevisePrompter(CreativePrompterBase):
  function load_answer (line 173) | def load_answer(tokenizer, cfg):
  function load_critique (line 182) | def load_critique(tokenizer, cfg):
  function load_revise (line 191) | def load_revise(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/dpo/chat_template.py
  function default (line 9) | def default(cfg, dataset_idx=0, **kwargs):
  function argilla_chat (line 125) | def argilla_chat(cfg, dataset_idx=0, **kwargs):

FILE: src/axolotl/prompt_strategies/dpo/chatml.py
  function default (line 6) | def default(
  function argilla_chat (line 46) | def argilla_chat(
  function icr (line 65) | def icr(
  function intel (line 91) | def intel(cfg, **kwargs):
  function prompt_pairs (line 113) | def prompt_pairs(cfg, **kwargs):
  function ultra (line 131) | def ultra(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/dpo/llama3.py
  function default (line 6) | def default(
  function argilla_chat (line 46) | def argilla_chat(
  function icr (line 65) | def icr(
  function intel (line 91) | def intel(cfg, **kwargs):
  function prompt_pairs (line 113) | def prompt_pairs(cfg, **kwargs):
  function ultra (line 131) | def ultra(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/dpo/passthrough.py
  function default (line 6) | def default(cfg, dataset_idx=0, **kwargs):

FILE: src/axolotl/prompt_strategies/dpo/user_defined.py
  function default (line 6) | def default(cfg, dataset_idx=0, **kwargs):

FILE: src/axolotl/prompt_strategies/dpo/zephyr.py
  function nectar (line 6) | def nectar(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/input_output.py
  class RawInputOutputStrategy (line 9) | class RawInputOutputStrategy(PromptTokenizingStrategy):
    method __init__ (line 12) | def __init__(self, *args, eos_token=None, **kwargs):
    method tokenize_prompt (line 18) | def tokenize_prompt(self, prompt):
  class RawInputOutputPrompter (line 40) | class RawInputOutputPrompter(Prompter):
    method build_prompt (line 43) | def build_prompt(self, source) -> Generator[Tuple[bool, str], None, No...
  function load (line 48) | def load(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/jinja_template_analyzer.py
  class JinjaTemplateAnalysis (line 9) | class JinjaTemplateAnalysis(TypedDict):
  class GenerationTagIgnore (line 31) | class GenerationTagIgnore(Extension):
    method parse (line 38) | def parse(self, parser):
  class JinjaTemplateAnalyzer (line 43) | class JinjaTemplateAnalyzer:
    method __init__ (line 72) | def __init__(self, template: str):
    method _visit_node (line 83) | def _visit_node(self, node) -> None:
    method _get_target_name (line 132) | def _get_target_name(self, node) -> Optional[str]:
    method _get_target_names (line 146) | def _get_target_names(self, node) -> list[str]:
    method _get_base_name (line 167) | def _get_base_name(self, node) -> Optional[str]:
    method get_template_variables (line 180) | def get_template_variables(self) -> Dict[str, Set[str]]:
    method analyze_template (line 212) | def analyze_template(self) -> Dict[str, JinjaTemplateAnalysis]:
    method get_downstream_properties (line 269) | def get_downstream_properties(self, start_var: str) -> Dict[str, Set[s...
    method get_message_vars (line 318) | def get_message_vars(self, field_messages: str = "messages") -> Set[str]:

FILE: src/axolotl/prompt_strategies/kto/chatml.py
  function argilla (line 6) | def argilla(
  function argilla_chat (line 26) | def argilla_chat(
  function intel (line 44) | def intel(cfg, **kwargs):
  function prompt_pairs (line 66) | def prompt_pairs(cfg, **kwargs):
  function ultra (line 83) | def ultra(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/kto/llama3.py
  function argilla (line 6) | def argilla(
  function argilla_chat (line 26) | def argilla_chat(
  function intel (line 44) | def intel(cfg, **kwargs):
  function prompt_pairs (line 66) | def prompt_pairs(cfg, **kwargs):
  function ultra (line 83) | def ultra(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/kto/user_defined.py
  function default (line 6) | def default(cfg, dataset_idx=0, **kwargs):

FILE: src/axolotl/prompt_strategies/llama2_chat.py
  class Llama2ChatConversation (line 38) | class Llama2ChatConversation:
    method get_prompt (line 58) | def get_prompt(self) -> str:
    method append_message (line 73) | def append_message(self, role: str, message: str):
  class LLama2ChatTokenizingStrategy (line 78) | class LLama2ChatTokenizingStrategy(PromptTokenizingStrategy):
    method __init__ (line 84) | def __init__(self, *args, **kwargs):
    method tokenize_prompt (line 91) | def tokenize_prompt(self, prompt):
  class Llama2ChatPrompter (line 156) | class Llama2ChatPrompter:
    method build_prompt (line 169) | def build_prompt(self, source) -> Generator[Llama2ChatConversation, No...
  function load (line 202) | def load(tokenizer, cfg) -> LLama2ChatTokenizingStrategy:

FILE: src/axolotl/prompt_strategies/messages/__init__.py
  function load (line 11) | def load(tokenizer, cfg, ds_cfg, processor=None):

FILE: src/axolotl/prompt_strategies/messages/chat.py
  class ChatMessageDatasetWrappingStrategy (line 12) | class ChatMessageDatasetWrappingStrategy(DatasetWrappingStrategy):
    method __init__ (line 17) | def __init__(
    method wrap_dataset (line 33) | def wrap_dataset(
  function load (line 51) | def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None):

FILE: src/axolotl/prompt_strategies/metharme.py
  class MetharmePromptTokenizingStrategy (line 14) | class MetharmePromptTokenizingStrategy(InstructionPromptTokenizingStrate...
    method parse_instruction_fields (line 19) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
    method _tokenize (line 22) | def _tokenize(
  class MetharmePrompter (line 56) | class MetharmePrompter(AlpacaPrompter):
    method __init__ (line 67) | def __init__(self, *args, **kwargs):
  function load (line 71) | def load(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/orcamini.py
  class OrcaMiniPrompter (line 19) | class OrcaMiniPrompter(AlpacaPrompter):
    method match_prompt_style (line 22) | def match_prompt_style(self):
    method build_prompt_w_system (line 27) | def build_prompt_w_system(
  function load (line 41) | def load(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/orpo/chat_template.py
  class Message (line 12) | class Message(BaseModel):
  class MessageList (line 20) | class MessageList(BaseModel):
  function load (line 26) | def load(tokenizer, cfg, ds_cfg: Optional[Dict[str, Any]] = None, **kwar...
  class ORPODatasetParsingStrategy (line 44) | class ORPODatasetParsingStrategy:
    method get_chosen_conversation_thread (line 47) | def get_chosen_conversation_thread(self, prompt) -> MessageList:
    method get_rejected_conversation_thread (line 63) | def get_rejected_conversation_thread(self, prompt) -> MessageList:
    method get_prompt (line 79) | def get_prompt(self, prompt) -> MessageList:
    method get_chosen (line 112) | def get_chosen(self, prompt) -> MessageList:
    method get_rejected (line 121) | def get_rejected(self, prompt) -> MessageList:
  class ORPOTokenizingStrategy (line 131) | class ORPOTokenizingStrategy(PromptTokenizingStrategy):
    method __init__ (line 141) | def __init__(
    method tokenize_prompt (line 150) | def tokenize_prompt(self, prompt):
  class ORPOPrompter (line 205) | class ORPOPrompter(Prompter):
    method __init__ (line 208) | def __init__(self, chat_template, tokenizer):
    method build_prompt (line 212) | def build_prompt(
  function argilla (line 251) | def argilla(cfg, **kwargs):

FILE: src/axolotl/prompt_strategies/pretrain.py
  class PretrainTokenizer (line 10) | class PretrainTokenizer:
    method build_prompt (line 13) | def build_prompt(self, prompt) -> Generator[str, None, None]:
  class PretrainTokenizationStrategy (line 17) | class PretrainTokenizationStrategy(PromptTokenizingStrategy):
    method supports_batched (line 21) | def supports_batched(self):
    method __init__ (line 24) | def __init__(self, *args, max_length=None, text_column="text", **kwargs):
    method _tokenize (line 30) | def _tokenize(
    method tokenize_prompt (line 48) | def tokenize_prompt(self, prompt):
  function load (line 52) | def load(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/pygmalion.py
  class PygmalionPromptTokenizingStrategy (line 19) | class PygmalionPromptTokenizingStrategy(PromptTokenizingStrategy):
    method __init__ (line 26) | def __init__(self, prompter, tokenizer, *args, **kwargs):
    method tokenize_prompt (line 31) | def tokenize_prompt(self, prompt):
  class PygmalionPrompter (line 82) | class PygmalionPrompter:
    method __init__ (line 87) | def __init__(self, *args, **kwargs):
    method build_prompt (line 90) | def build_prompt(
  function load (line 100) | def load(tokenizer, cfg):

FILE: src/axolotl/prompt_strategies/stepwise_supervised.py
  class StepwiseSupervisedPromptTokenizingStrategy (line 15) | class StepwiseSupervisedPromptTokenizingStrategy:
    method __init__ (line 24) | def __init__(
    method tokenize_prompt (line 38) | def tokenize_prompt(
    method supports_batched (line 101) | def supports_batched(self):
  function load (line 105) | def load(

FILE: src/axolotl/prompt_strategies/user_defined.py
  class UserDefinedDatasetConfig (line 16) | class UserDefinedDatasetConfig:
    method __getitem__ (line 30) | def __getitem__(self, item):
  class UserDefinedPromptTokenizationStrategy (line 34) | class UserDefinedPromptTokenizationStrategy(InstructionWSystemPromptToke...
  function load (line 40) | def load(tokenizer, cfg, ds_cfg: Optional[UserDefinedDatasetConfig] = No...

FILE: src/axolotl/prompt_tokenizers.py
  class InvalidDataException (line 21) | class InvalidDataException(Exception):
  class DatasetWrappingStrategy (line 27) | class DatasetWrappingStrategy(abc.ABC):
    method wrap_dataset (line 33) | def wrap_dataset(
  class PromptTokenizingStrategy (line 43) | class PromptTokenizingStrategy(abc.ABC):
    method __init__ (line 50) | def __init__(
    method tokenize_prompt (line 66) | def tokenize_prompt(self, prompt):
    method supports_batched (line 70) | def supports_batched(self):
    method _tokenize (line 73) | def _tokenize(
  class InstructionPromptTokenizingStrategy (line 108) | class InstructionPromptTokenizingStrategy(PromptTokenizingStrategy):
    method parse_instruction_fields (line 113) | def parse_instruction_fields(
    method tokenize_prompt (line 118) | def tokenize_prompt(self, prompt):
    method _build_full_prompt (line 146) | def _build_full_prompt(
  class AlpacaPromptTokenizingStrategy (line 163) | class AlpacaPromptTokenizingStrategy(InstructionPromptTokenizingStrategy):
    method parse_instruction_fields (line 168) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class AlpacaMultipleChoicePromptTokenizingStrategy (line 176) | class AlpacaMultipleChoicePromptTokenizingStrategy(InstructionPromptToke...
    method parse_instruction_fields (line 181) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class JeopardyPromptTokenizingStrategy (line 189) | class JeopardyPromptTokenizingStrategy(InstructionPromptTokenizingStrate...
    method parse_instruction_fields (line 194) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class OpenAssistantPromptTokenizingStrategy (line 202) | class OpenAssistantPromptTokenizingStrategy(InstructionPromptTokenizingS...
    method parse_instruction_fields (line 207) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class SummarizeTLDRPromptTokenizingStrategy (line 215) | class SummarizeTLDRPromptTokenizingStrategy(InstructionPromptTokenizingS...
    method parse_instruction_fields (line 220) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class GPTeacherPromptTokenizingStrategy (line 228) | class GPTeacherPromptTokenizingStrategy(InstructionPromptTokenizingStrat...
    method parse_instruction_fields (line 233) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class NomicGPT4AllPromptTokenizingStrategy (line 241) | class NomicGPT4AllPromptTokenizingStrategy(InstructionPromptTokenizingSt...
    method parse_instruction_fields (line 246) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str]:
  class ReflectionPromptTokenizingStrategy (line 254) | class ReflectionPromptTokenizingStrategy(PromptTokenizingStrategy):
    method parse_instruction_fields (line 259) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str, str...
    method tokenize_prompt (line 262) | def tokenize_prompt(self, prompt):
    method _build_full_prompt (line 292) | def _build_full_prompt(self, instruction, input, output, reflection, c...
    method _tokenize (line 305) | def _tokenize(self, prompt, add_eos_token=True, strip_bos_token=False):
  class AlpacaReflectionPTStrategy (line 325) | class AlpacaReflectionPTStrategy(ReflectionPromptTokenizingStrategy):
    method parse_instruction_fields (line 330) | def parse_instruction_fields(self, prompt) -> Tuple[str, str, str, str...
  function tokenize_prompt_default (line 340) | def tokenize_prompt_default() -> Tuple[Dict[str, List[int]], int]:
  function parse_tokenized_to_result (line 354) | def parse_tokenized_to_result(

FILE: src/axolotl/prompters.py
  class PromptStyle (line 15) | class PromptStyle(Enum):
  class Prompter (line 26) | class Prompter:
  class AlpacaPrompter (line 32) | class AlpacaPrompter(Prompter):
    method __init__ (line 44) | def __init__(self, prompt_style: Optional[str] = PromptStyle.INSTRUCT....
    method match_prompt_style (line 48) | def match_prompt_style(self):
    method _build_result (line 72) | def _build_result(self, instruction, input_text, output):
    method build_prompt (line 92) | def build_prompt(
    method __repr__ (line 100) | def __repr__(self) -> str:
  class UnpromptedPrompter (line 106) | class UnpromptedPrompter(AlpacaPrompter):
  class JeopardyPrompter (line 115) | class JeopardyPrompter(AlpacaPrompter):
  class MultipleChoiceExplainPrompter (line 123) | class MultipleChoiceExplainPrompter(AlpacaPrompter):
  class MultipleChoiceConcisePrompter (line 136) | class MultipleChoiceConcisePrompter(AlpacaPrompter):
    method match_prompt_style (line 144) | def match_prompt_style(self):
  class SummarizeTLDRPrompter (line 149) | class SummarizeTLDRPrompter(AlpacaPrompter):
    method match_prompt_style (line 157) | def match_prompt_style(self):
  class GPTeacherPrompter (line 162) | class GPTeacherPrompter(AlpacaPrompter):
  class NomicGPT4AllPrompter (line 168) | class NomicGPT4AllPrompter(AlpacaPrompter):
  class ReflectAlpacaPrompter (line 174) | class ReflectAlpacaPrompter(Prompter):
    method __init__ (line 189) | def __init__(self, prompt_style="instruct"):
    method match_prompt_style (line 193) | def match_prompt_style(self):
    method _build_result (line 217) | def _build_result(
    method build_prompt (line 241) | def build_prompt(
    method __repr__ (line 257) | def __repr__(self) -> str:
  class UnsupportedPrompter (line 268) | class UnsupportedPrompter(Prompter):
    method __init__ (line 273) | def __init__(self) -> None:
    method __repr__ (line 276) | def __repr__(self):

FILE: src/axolotl/scripts/vllm_serve_lora.py
  class LoRAScriptArguments (line 38) | class LoRAScriptArguments(ScriptArguments):
  function llm_worker (line 59) | def llm_worker(
  function main (line 125) | def main(script_args: ScriptArguments):

FILE: src/axolotl/scripts/vllm_worker_ext.py
  class BatchWeightSyncWorkerExtension (line 33) | class BatchWeightSyncWorkerExtension(WeightSyncWorkerExtension):
    method init_communicator (line 36) | def init_communicator(self, host, port, world_size, client_device_uuid):
    method _direct_set_weight (line 42) | def _direct_set_weight(self, name: str, weight: torch.Tensor) -> None:
    method update_named_param (line 112) | def update_named_param(self, name, dtype, shape):
    method batch_update_named_params (line 129) | def batch_update_named_params(self, params_list: list[tuple[str, str, ...

FILE: src/axolotl/telemetry/callbacks.py
  class TelemetryCallback (line 21) | class TelemetryCallback(TrainerCallback):
    method __init__ (line 31) | def __init__(self):
    method on_train_begin (line 41) | def on_train_begin(
    method on_train_end (line 52) | def on_train_end(
    method on_epoch_begin (line 68) | def on_epoch_begin(
    method on_epoch_end (line 80) | def on_epoch_end(
    method on_step_end (line 91) | def on_step_end(
    method _extract_last_metrics (line 155) | def _extract_last_metrics(self, state: TrainerState) -> dict:

FILE: src/axolotl/telemetry/errors.py
  function sanitize_stack_trace (line 18) | def sanitize_stack_trace(stack_trace: str) -> str:
  function send_errors (line 104) | def send_errors(func: Callable) -> Callable:

FILE: src/axolotl/telemetry/manager.py
  function is_main_process (line 61) | def is_main_process() -> bool:
  class TelemetryManager (line 98) | class TelemetryManager:
    method __new__ (line 104) | def __new__(cls):
    method __init__ (line 115) | def __init__(self):
    method get_instance (line 140) | def get_instance(cls) -> "TelemetryManager":
    method _check_telemetry_enabled (line 146) | def _check_telemetry_enabled(self) -> bool:
    method _load_whitelist (line 187) | def _load_whitelist(self) -> dict:
    method _is_whitelisted (line 199) | def _is_whitelisted(self, value: str) -> bool:
    method _init_posthog (line 220) | def _init_posthog(self):
    method _redact_paths (line 226) | def _redact_paths(self, properties: dict[str, Any]) -> dict[str, Any]:
    method _get_system_info (line 267) | def _get_system_info(self) -> dict[str, Any]:
    method send_event (line 359) | def send_event(self, event_type: str, properties: dict[str, Any] | Non...
    method send_system_info (line 387) | def send_system_info(self):
    method shutdown (line 392) | def shutdown(self):

FILE: src/axolotl/telemetry/runtime_metrics.py
  class RuntimeMetrics (line 17) | class RuntimeMetrics:
    method __post_init__ (line 34) | def __post_init__(self):
    method elapsed_time (line 41) | def elapsed_time(self) -> float:
    method epoch_time (line 45) | def epoch_time(self, epoch: int) -> float | None:
    method average_epoch_time (line 52) | def average_epoch_time(self) -> float | None:
    method steps_per_second (line 68) | def steps_per_second(self) -> float | None:
    method to_dict (line 75) | def to_dict(self) -> dict[str, Any]:
  class RuntimeMetricsTracker (line 112) | class RuntimeMetricsTracker:
    method __init__ (line 117) | def __init__(self):
    method start_epoch (line 123) | def start_epoch(self, epoch: int):
    method end_epoch (line 129) | def end_epoch(self, epoch: int):
    method update_step (line 133) | def update_step(self, step: int):
    method _get_allocated_memory (line 142) | def _get_allocated_memory(self) -> dict[int, int]:
    method update_memory_metrics (line 182) | def update_memory_metrics(self):
    method get_memory_metrics (line 195) | def get_memory_metrics(self) -> dict[str, Any]:

FILE: src/axolotl/train.py
  function setup_model_and_tokenizer (line 54) | def setup_model_and_tokenizer(
  function setup_reference_model (line 120) | def setup_reference_model(
  function setup_signal_handler (line 149) | def setup_signal_handler(cfg: DictDefault, model: PreTrainedModel):
  function execute_training (line 175) | def execute_training(
  function save_trained_model (line 224) | def save_trained_model(
  function create_model_card (line 354) | def create_model_card(cfg: DictDefault, trainer: Trainer):
  function save_initial_configs (line 390) | def save_initial_configs(
  function setup_model_card (line 430) | def setup_model_card(cfg: DictDefault):
  function handle_untrained_tokens_fix (line 447) | def handle_untrained_tokens_fix(
  function setup_model_and_trainer (line 487) | def setup_model_and_trainer(
  function train (line 558) | def train(

FILE: src/axolotl/utils/__init__.py
  function is_mlflow_available (line 12) | def is_mlflow_available():
  function is_comet_available (line 16) | def is_comet_available():
  function is_opentelemetry_available (line 20) | def is_opentelemetry_available():
  function is_trackio_available (line 27) | def is_trackio_available():
  function get_pytorch_version (line 31) | def get_pytorch_version() -> tuple[int, int, int]:
  function set_pytorch_cuda_alloc_conf (line 47) | def set_pytorch_cuda_alloc_conf():
  function set_misc_env (line 66) | def set_misc_env():
  function get_not_null (line 71) | def get_not_null(value, default=None):

FILE: src/axolotl/utils/bench.py
  function check_cuda_device (line 25) | def check_cuda_device(default_value):
  function gpu_memory_usage (line 54) | def gpu_memory_usage(device=0):
  function gpu_memory_usage_all (line 59) | def gpu_memory_usage_all(device=0):
  function mps_memory_usage_all (line 67) | def mps_memory_usage_all():
  function npu_memory_usage_all (line 73) | def npu_memory_usage_all(device=0):
  function gpu_memory_usage_smi (line 80) | def gpu_memory_usage_smi(device=0):
  function get_gpu_memory_usage (line 96) | def get_gpu_memory_usage(device: int | torch.device = 0):
  function log_gpu_memory_usage (line 110) | def log_gpu_memory_usage(

FILE: src/axolotl/utils/callbacks/__init__.py
  class LossWatchDogCallback (line 57) | class LossWatchDogCallback(TrainerCallback):
    method __init__ (line 60) | def __init__(self, cfg):
    method on_step_end (line 66) | def on_step_end(
  class SaveModelOnFirstStepCallback (line 86) | class SaveModelOnFirstStepCallback(TrainerCallback):
    method on_step_end (line 89) | def on_step_end(
  function bench_eval_callback_factory (line 101) | def bench_eval_callback_factory(trainer, tokenizer):
  function causal_lm_bench_eval_callback_factory (line 303) | def causal_lm_bench_eval_callback_factory(trainer: Trainer, tokenizer):
  function log_prediction_callback_factory (line 512) | def log_prediction_callback_factory(trainer: Trainer, tokenizer, logger:...
  class SaveAxolotlConfigtoWandBCallback (line 731) | class SaveAxolotlConfigtoWandBCallback(TrainerCallback):
    method __init__ (line 734) | def __init__(self, axolotl_config_path):
    method on_train_begin (line 737) | def on_train_begin(
  class GCCallback (line 829) | class GCCallback(TrainerCallback):
    method __init__ (line 832) | def __init__(self, gc_steps: int | None = -1):
    method _gc (line 836) | def _gc(self):
    method on_train_begin (line 840) | def on_train_begin(
    method on_step_begin (line 849) | def on_step_begin(
    method on_step_end (line 859) | def on_step_end(
    method on_epoch_end (line 885) | def on_epoch_end(
  function colab_inference_post_train_callback (line 895) | def colab_inference_post_train_callback(trainer: Trainer):

FILE: src/axolotl/utils/callbacks/comet_.py
  class SaveAxolotlConfigtoCometCallback (line 17) | class SaveAxolotlConfigtoCometCallback(TrainerCallback):
    method __init__ (line 20) | def __init__(self, axolotl_config_path):
    method on_train_begin (line 23) | def on_train_begin(

FILE: src/axolotl/utils/callbacks/dynamic_checkpoint.py
  class DynamicCheckpointCallback (line 22) | class DynamicCheckpointCallback(TrainerCallback):
    method _get_config_value (line 34) | def _get_config_value(self, config, key, default=None):
    method __init__ (line 40) | def __init__(self, cfg):
    method on_step_end (line 64) | def on_step_end(

FILE: src/axolotl/utils/callbacks/generation.py
  class SFTGenerationCallback (line 12) | class SFTGenerationCallback(TrainerCallback):
    method __init__ (line 15) | def __init__(self, trainer):
    method on_evaluate (line 18) | def on_evaluate(
    method _log_samples (line 62) | def _log_samples(self, samples: list, step: int):

FILE: src/axolotl/utils/callbacks/lisa.py
  function lisa_callback_factory (line 23) | def lisa_callback_factory(trainer: "AxolotlTrainer"):

FILE: src/axolotl/utils/callbacks/mlflow_.py
  function should_log_artifacts (line 20) | def should_log_artifacts() -> bool:
  class SaveAxolotlConfigtoMlflowCallback (line 25) | class SaveAxolotlConfigtoMlflowCallback(TrainerCallback):
    method __init__ (line 28) | def __init__(self, axolotl_config_path):
    method on_train_begin (line 31) | def on_train_begin(

FILE: src/axolotl/utils/callbacks/models.py
  function get_causal_lm_model_cls_prefix (line 8) | def get_causal_lm_model_cls_prefix(model_type: str) -> Tuple[str, str]:

FILE: src/axolotl/utils/callbacks/opentelemetry.py
  class OpenTelemetryMetricsCallback (line 30) | class OpenTelemetryMetricsCallback(TrainerCallback):
    method __init__ (line 45) | def __init__(self, cfg):
    method _create_metrics (line 76) | def _create_metrics(self):
    method _start_metrics_server (line 120) | def _start_metrics_server(self):
    method on_train_begin (line 135) | def on_train_begin(
    method on_log (line 149) | def on_log(
    method on_step_end (line 178) | def on_step_end(
    method on_evaluate (line 194) | def on_evaluate(
    method on_train_end (line 224) | def on_train_end(

FILE: src/axolotl/utils/callbacks/perplexity.py
  class Perplexity (line 19) | class Perplexity:
    method __init__ (line 25) | def __init__(
    method _feature_names (line 36) | def _feature_names(self) -> List[str]:
    method compute (line 39) | def compute(

FILE: src/axolotl/utils/callbacks/profiler.py
  class PytorchProfilerCallback (line 17) | class PytorchProfilerCallback(TrainerCallback):
    method __init__ (line 24) | def __init__(self, steps_to_profile: int = 5, profiler_steps_start: in...
    method on_step_begin (line 36) | def on_step_begin(
    method on_step_end (line 60) | def on_step_end(
    method on_train_end (line 82) | def on_train_end(

FILE: src/axolotl/utils/callbacks/qat.py
  function toggle_fake_quant (line 16) | def toggle_fake_quant(mod: nn.Module, enable: bool):
  class QATCallback (line 33) | class QATCallback(TrainerCallback):
    method __init__ (line 38) | def __init__(self, cfg: QATConfig):
    method on_step_begin (line 41) | def on_step_begin(self, args, state, control, model, **kwargs):

FILE: src/axolotl/utils/callbacks/swanlab.py
  class CustomSwanLabCallback (line 26) | class CustomSwanLabCallback(TrainerCallback):
    method __init__ (line 34) | def __init__(self):
    method setup (line 38) | def setup(self):
    method on_train_begin (line 58) | def on_train_begin(
    method on_log (line 95) | def on_log(
    method on_train_end (line 129) | def on_train_end(
  class SaveAxolotlConfigtoSwanLabCallback (line 146) | class SaveAxolotlConfigtoSwanLabCallback(TrainerCallback):
    method __init__ (line 149) | def __init__(self, axolotl_config_path):
    method on_train_begin (line 152) | def on_train_begin(

FILE: src/axolotl/utils/callbacks/tokens_per_second.py
  class TokensPerSecondCallback (line 22) | class TokensPerSecondCallback(TrainerCallback):
    method __init__ (line 28) | def __init__(
    method on_train_begin (line 41) | def on_train_begin(
    method on_step_begin (line 61) | def on_step_begin(
    method on_step_end (line 73) | def on_step_end(
    method on_log (line 95) | def on_log(

FILE: src/axolotl/utils/callbacks/trackio_.py
  class SaveAxolotlConfigtoTrackioCallback (line 18) | class SaveAxolotlConfigtoTrackioCallback(TrainerCallback):
    method __init__ (line 21) | def __init__(self, axolotl_config_path):
    method on_train_begin (line 24) | def on_train_begin(

FILE: src/axolotl/utils/chat_templates/base.py
  function get_chat_template (line 26) | def get_chat_template(
  function extract_chat_template_args (line 88) | def extract_chat_template_args(cfg, ds_cfg: Dict[str, Any] | None = None):
  function get_chat_template_from_config (line 98) | def get_chat_template_from_config(
  function register_chat_template (line 113) | def register_chat_template(template_name: str, chat_template: str):

FILE: src/axolotl/utils/collators/batching.py
  class DataCollatorForSeq2Seq (line 12) | class DataCollatorForSeq2Seq:
    method __call__ (line 55) | def __call__(self, features, return_tensors=None):
  class BatchSamplerDataCollatorForSeq2Seq (line 129) | class BatchSamplerDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
    method __call__ (line 134) | def __call__(self, features, return_tensors=None):
  class V2BatchSamplerDataCollatorForSeq2Seq (line 159) | class V2BatchSamplerDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
    method __call__ (line 166) | def __call__(self, features, return_tensors=None):
  class PretrainingBatchSamplerDataCollatorForSeq2Seq (line 200) | class PretrainingBatchSamplerDataCollatorForSeq2Seq(DataCollatorForSeq2S...
    method __init__ (line 205) | def __init__(self, *args, multipack_attn=True, **kwargs):
    method __call__ (line 209) | def __call__(self, features, return_tensors=None):

FILE: src/axolotl/utils/collators/mamba.py
  class MambaDataCollator (line 15) | class MambaDataCollator:
    method __call__ (line 22) | def __call__(self, instances: Sequence[Dict]) -> Dict[str, torch.Tensor]:

FILE: src/axolotl/utils/collators/mm_chat.py
  class MultiModalChatDataCollator (line 17) | class MultiModalChatDataCollator(DataCollatorMixin):
    method __post_init__ (line 29) | def __post_init__(self):
    method torch_call (line 33) | def torch_call(self, examples: list[dict]) -> dict[str, Any]:
    method process_rows (line 36) | def process_rows(

FILE: src/axolotl/utils/comet_.py
  function python_value_to_environ_value (line 45) | def python_value_to_environ_value(python_value):
  function setup_comet_env_vars (line 61) | def setup_comet_env_vars(cfg: DictDefault):

FILE: src/axolotl/utils/config/__init__.py
  function choose_device (line 30) | def choose_device(cfg):
  function resolve_dtype (line 64) | def resolve_dtype(cfg):
  function normalize_config (line 107) | def normalize_config(cfg):
  function normalize_cfg_datasets (line 269) | def normalize_cfg_datasets(cfg):
  function validate_config (line 288) | def validate_config(
  function prepare_plugins (line 337) | def prepare_plugins(cfg):

FILE: src/axolotl/utils/ctx_managers/sequence_parallel.py
  function apply_sequence_parallelism (line 24) | def apply_sequence_parallelism(
  class SequenceParallelContextManager (line 170) | class SequenceParallelContextManager:
    method __init__ (line 189) | def __init__(
    method __enter__ (line 233) | def __enter__(self):
    method __exit__ (line 238) | def __exit__(self, exc_type, exc_val, exc_tb):
    method _register_ring_attn (line 246) | def _register_ring_attn(self):
    method _register_model_hooks (line 255) | def _register_model_hooks(self):
    method _gather_outputs (line 359) | def _gather_outputs(self, output: CausalLMOutputWithPast) -> CausalLMO...
  class AllGatherWithGrad (line 368) | class AllGatherWithGrad(torch.autograd.Function):
    method forward (line 372) | def forward(
    method backward (line 419) | def backward(

FILE: src/axolotl/utils/data/lock.py
  class FileLockLoader (line 17) | class FileLockLoader:
    method __init__ (line 24) | def __init__(self, cfg: DictDefault):
    method load (line 33) | def load(self, load_fn: Callable[[], Any]) -> Any:
    method _increment_counter (line 46) | def _increment_counter(self):
    method cleanup (line 55) | def cleanup(self):

FILE: src/axolotl/utils/data/rl.py
  function prepare_preference_datasets (line 37) | def prepare_preference_datasets(
  function _map_dataset (line 86) | def _map_dataset(
  function _drop_long_sequences (line 125) | def _drop_long_sequences(
  function _load_split (line 182) | def _load_split(cfg: DictDefault, split: Literal["train", "test"]) -> Da...
  function _load_or_create_dataset_split (line 262) | def _load_or_create_dataset_split(

FILE: src/axolotl/utils/data/sft.py
  function prepare_datasets (line 48) | def prepare_datasets(
  function _prepare_standard_dataset (line 68) | def _prepare_standard_dataset(
  function _prepare_streaming_dataset (line 125) | def _prepare_streaming_dataset(
  function _extract_pretraining_config (line 176) | def _extract_pretraining_config(cfg: DictDefault) -> DictDefault:
  function _load_streaming_dataset (line 205) | def _load_streaming_dataset(
  function _create_placeholder_dataset (line 251) | def _create_placeholder_dataset() -> IterableDataset:
  function _load_tokenized_prepared_datasets (line 260) | def _load_tokenized_prepared_datasets(
  function _load_raw_datasets (line 311) | def _load_raw_datasets(
  function _load_and_process_single_dataset (line 369) | def _load_and_process_single_dataset(
  function _parse_dataset_type (line 420) | def _parse_dataset_type(d_type: str) -> tuple[str | None, str | None]:
  function _handle_train_dataset_split (line 432) | def _handle_train_dataset_split(
  function _apply_dataset_sharding (line 451) | def _apply_dataset_sharding(dataset: Dataset, cfg: DictDefault) -> Dataset:
  function _load_and_prepare_datasets (line 472) | def _load_and_prepare_datasets(

FILE: src/axolotl/utils/data/shared.py
  function get_dataset_type (line 48) | def get_dataset_type(dataset_config: DictDefault) -> str:
  function datasets_with_name_generator (line 60) | def datasets_with_name_generator(
  function load_dataset_with_config (line 93) | def load_dataset_with_config(
  function _check_if_hub_dataset (line 151) | def _check_if_hub_dataset(dataset_config: DictDefault, use_auth_token: b...
  function _get_remote_filesystem (line 173) | def _get_remote_filesystem(
  function _load_from_local_path (line 222) | def _load_from_local_path(
  function _load_from_hub (line 258) | def _load_from_hub(
  function _load_from_cloud (line 271) | def _load_from_cloud(
  function _load_from_url (line 298) | def _load_from_url(
  function _load_from_data_files (line 310) | def _load_from_data_files(
  function generate_split_fingerprints (line 339) | def generate_split_fingerprints(
  function get_prepared_dataset_path (line 354) | def get_prepared_dataset_path(cfg: DictDefault, dataset_hash: str) -> Path:
  function create_train_validation_split (line 368) | def create_train_validation_split(
  function _generate_from_iterable_dataset (line 400) | def _generate_from_iterable_dataset(
  function save_preprocessed_dataset (line 409) | def save_preprocessed_dataset(
  function load_preprocessed_dataset (line 457) | def load_preprocessed_dataset(cfg: DictDefault, dataset_hash: str) -> Da...
  function try_load_from_hub (line 486) | def try_load_from_hub(
  function generate_dataset_hash_from_config (line 506) | def generate_dataset_hash_from_config(
  function merge_datasets (line 529) | def merge_datasets(datasets: list[Dataset], cfg: DictDefault) -> Dataset:

FILE: src/axolotl/utils/data/streaming.py
  function encode_streaming (line 20) | def encode_streaming(
  function wrap_streaming_dataset (line 179) | def wrap_streaming_dataset(
  function encode_packed_streaming (line 254) | def encode_packed_streaming(

FILE: src/axolotl/utils/data/utils.py
  class RetryStrategy (line 23) | class RetryStrategy(Enum):
  function retry_on_request_exceptions (line 31) | def retry_on_request_exceptions(
  function md5 (line 73) | def md5(to_hash: str, encoding: str = "utf-8") -> str:
  function sha256 (line 81) | def sha256(to_hash: str, encoding: str = "utf-8") -> str:
  function _deduplicate_dataset (line 86) | def _deduplicate_dataset(
  function deduplicate_and_log_datasets (line 112) | def deduplicate_and_log_datasets(
  function keep_min_len (line 151) | def keep_min_len(sample, min_sequence_len=2):
  function truncate_long_seq (line 167) | def truncate_long_seq(sample, sequence_len=2048):
  function _should_skip_processing (line 188) | def _should_skip_processing(dataset: Dataset) -> bool:
  function _log_dataset_stats (line 208) | def _log_dataset_stats(dataset: Dataset) -> None:
  function _build_filter_kwargs (line 216) | def _build_filter_kwargs(dataset: Dataset, cfg: DictDefault) -> dict:
  function _filter_short_sequences (line 225) | def _filter_short_sequences(
  function _truncate_long_sequences (line 251) | def _truncate_long_sequences(
  function _drop_outside_range (line 269) | def _drop_outside_range(
  function handle_long_seq_in_dataset (line 314) | def handle_long_seq_in_dataset(

FILE: src/axolotl/utils/data/wrappers.py
  function handle_unknown_dataset_strategy (line 45) | def handle_unknown_dataset_strategy(dataset_config: DictDefault) -> NoRe...
  function get_dataset_wrapper (line 57) | def get_dataset_wrapper(
  function _is_dataset_already_tokenized (line 134) | def _is_dataset_already_tokenized(dataset: Dataset | IterableDataset) ->...
  function _handle_custom_dataset_type (line 144) | def _handle_custom_dataset_type(
  function _handle_bradley_terry_dataset (line 165) | def _handle_bradley_terry_dataset(
  function _handle_stepwise_supervised_dataset (line 189) | def _handle_stepwise_supervised_dataset(
  function _handle_loaded_strategy (line 213) | def _handle_loaded_strategy(
  function _handle_alpaca_dataset (line 231) | def _handle_alpaca_dataset(
  function _handle_explainchoice_dataset (line 254) | def _handle_explainchoice_dataset(
  function _handle_concisechoice_dataset (line 277) | def _handle_concisechoice_dataset(
  function _handle_summarizetldr_dataset (line 300) | def _handle_summarizetldr_dataset(
  function _handle_jeopardy_dataset (line 323) | def _handle_jeopardy_dataset(
  function _handle_oasst_dataset (line 346) | def _handle_oasst_dataset(
  function _handle_gpteacher_dataset (line 369) | def _handle_gpteacher_dataset(
  function _handle_reflection_dataset (line 392) | def _handle_reflection_dataset(

FILE: src/axolotl/utils/datasets.py
  function get_default_process_count (line 10) | def get_default_process_count():

FILE: src/axolotl/utils/dict.py
  class DictDefault (line 6) | class DictDefault(Dict):
    method __missing__ (line 11) | def __missing__(self, key):
    method __or__ (line 14) | def __or__(self, other):
    method __setitem__ (line 17) | def __setitem__(self, name, value):
  function remove_none_values (line 41) | def remove_none_values(obj):

FILE: src/axolotl/utils/distributed.py
  function get_device_type (line 21) | def get_device_type() -> torch.device:
  function get_device_count (line 32) | def get_device_count() -> int:
  function get_current_device (line 41) | def get_current_device() -> int:
  function init_distributed_state (line 50) | def init_distributed_state():
  function get_distributed_state (line 60) | def get_distributed_state() -> PartialState | None:
  function is_distributed (line 64) | def is_distributed() -> bool:
  function barrier (line 74) | def barrier():
  function is_main_process (line 83) | def is_main_process() -> bool:
  function is_local_main_process (line 101) | def is_local_main_process() -> bool:
  function get_world_size (line 107) | def get_world_size() -> int:
  function cleanup_distributed (line 111) | def cleanup_distributed():
  function zero_first (line 129) | def zero_first(is_main: bool):
  function gather_scalar_from_all_ranks (line 140) | def gather_scalar_from_all_ranks(fn, world_size=1):
  function broadcast_dict (line 176) | def broadcast_dict(vals: dict):
  function compute_and_broadcast (line 204) | def compute_and_broadcast(fn):
  function gather_from_all_ranks (line 237) | def gather_from_all_ranks(fn, world_size=1):
  function reduce_and_broadcast (line 274) | def reduce_and_broadcast(fn1, fn2):
  function build_parallelism_config (line 299) | def build_parallelism_config(cfg):
  function _get_parallel_config_kwargs (line 319) | def _get_parallel_config_kwargs(

FILE: src/axolotl/utils/environment.py
  function check_cuda_p2p_ib_support (line 19) | def check_cuda_p2p_ib_support():
  function check_cuda_p2p_support (line 27) | def check_cuda_p2p_support() -> bool:
  function get_package_version (line 49) | def get_package_version(package: str) -> Version:
  function is_package_version_ge (line 54) | def is_package_version_ge(package: str, version_: str) -> bool:

FILE: src/axolotl/utils/freeze.py
  function freeze_layers_except (line 14) | def freeze_layers_except(model, regex_patterns):
  function _invert_ranges (line 72) | def _invert_ranges(
  function _merge_ranges (line 102) | def _merge_ranges(
  function _create_freeze_parameters_hook (line 145) | def _create_freeze_parameters_hook(ranges_to_freeze: List[Tuple[int, int...
  class LayerNamePattern (line 173) | class LayerNamePattern:
    method __init__ (line 178) | def __init__(self, pattern: str):
    method match (line 189) | def match(self, name: str) -> bool:
    method _parse_pattern (line 201) | def _parse_pattern(

FILE: src/axolotl/utils/generation/sft.py
  function generate_samples (line 14) | def generate_samples(
  function format_generation_for_logging (line 143) | def format_generation_for_logging(

FILE: src/axolotl/utils/import_helper.py
  function get_cls_from_module_str (line 8) | def get_cls_from_module_str(module_str: str):

FILE: src/axolotl/utils/logging.py
  class MultiProcessAdapter (line 20) | class MultiProcessAdapter(logging.LoggerAdapter):
    method _should_log (line 26) | def _should_log(main_process_only: bool):
    method log (line 29) | def log(self, level, msg, *args, **kwargs):
    method warning_once (line 38) | def warning_once(self, *args, **kwargs):
  function get_logger (line 49) | def get_logger(name: str, log_level: str | None = None) -> MultiProcessA...

FILE: src/axolotl/utils/lora.py
  function get_lora_merged_state_dict (line 24) | def get_lora_merged_state_dict(

FILE: src/axolotl/utils/mistral/mistral3_processor.py
  class Mistral3ProcessorKwargs (line 14) | class Mistral3ProcessorKwargs(ProcessingKwargs):
  class Mistral3Processor (line 27) | class Mistral3Processor(ProcessorMixin):
    method __init__ (line 33) | def __init__(self, tokenizer: HFMistralTokenizer):
    method audio_tokenizer (line 37) | def audio_tokenizer(self) -> None:
    method _merge_kwargs (line 41) | def _merge_kwargs(
    method apply_chat_template (line 64) | def apply_chat_template(
    method __call__ (line 140) | def __call__(

FILE: src/axolotl/utils/mistral/mistral_tokenizer.py
  class HFMistralTokenizer (line 14) | class HFMistralTokenizer(MistralCommonBackend):
    method __init__ (line 20) | def __init__(self, name_or_path: str, **kwargs):
    method name_or_path (line 37) | def name_or_path(self) -> str:
    method name_or_path (line 41) | def name_or_path(self, name_or_path: str) -> None:
    method chat_template (line 45) | def chat_template(self) -> str | None:
    method chat_template (line 50) | def chat_template(self, chat_template: str | None) -> None:
    method _set_mode (line 53) | def _set_mode(self, mode: ValidationMode):
    method apply_chat_template (line 82) | def apply_chat_template(  # type: ignore
    method decode (line 107) | def decode(  # type: ignore
    method from_pretrained (line 124) | def from_pretrained(
    method save_pretrained (line 233) | def save_pretrained(self, *args, **kwargs) -> tuple[str, ...]:

FILE: src/axolotl/utils/mlflow_.py
  function setup_mlflow_env_vars (line 8) | def setup_mlflow_env_vars(cfg: DictDefault):

FILE: src/axolotl/utils/model_shard_quant.py
  function _replace_linear (line 21) | def _replace_linear(
  function load_and_quantize (line 62) | def load_and_quantize(
  function n_loading_workers (line 128) | def n_loading_workers(quant_method: str, param_count: float):
  function load_sharded_model (line 140) | def load_sharded_model(
  function load_sharded_model_quant (line 167) | def load_sharded_model_quant(

FILE: src/axolotl/utils/optimizers/adopt.py
  class ADOPT (line 39) | class ADOPT(Optimizer):
    method __init__ (line 40) | def __init__(
    method __setstate__ (line 104) | def __setstate__(self, state):
    method _init_group (line 126) | def _init_group(
    method step (line 192) | def step(self, closure=None):
  function _single_tensor_adopt (line 249) | def _single_tensor_adopt(
  function _multi_tensor_adopt (line 331) | def _multi_tensor_adopt(
  function adopt (line 460) | def adopt(

FILE: src/axolotl/utils/quantization.py
  function get_quantization_config (line 52) | def get_quantization_config(
  function quantize_model (line 138) | def quantize_model(
  function prepare_model_for_qat (line 176) | def prepare_model_for_qat(
  function convert_qat_model (line 231) | def convert_qat_model(

FILE: src/axolotl/utils/samplers/multipack.py
  function ffd_check (line 25) | def ffd_check(sequence_lengths: np.ndarray, bin_capacity: int, num_bins:...
  function pack_group (line 61) | def pack_group(
  function _process_group (line 115) | def _process_group(
  function pack_parallel (line 125) | def pack_parallel(
  function allocate_sequentially (line 194) | def allocate_sequentially(
  class MultipackBatchSampler (line 244) | class MultipackBatchSampler(BatchSampler):
    method __init__ (line 257) | def __init__(
    method set_epoch (line 305) | def set_epoch(self, epoch: int):
    method generate_batches (line 310) | def generate_batches(self, set_stats: bool = False) -> list[list[list[...
    method __iter__ (line 383) | def __iter__(self) -> Iterator[list[list[int]]]:
    method efficiency (line 395) | def efficiency(self) -> float:
    method gather_efficiency (line 406) | def gather_efficiency(self) -> float:
    method gather_len_batches (line 432) | def gather_len_batches(self, num: int) -> int:
    method __len__ (line 445) | def __len__(self) -> int:

FILE: src/axolotl/utils/samplers/utils.py
  function get_dataset_lengths (line 8) | def get_dataset_lengths(dataset, from_arrow=False):

FILE: src/axolotl/utils/schedulers.py
  class RexLR (line 12) | class RexLR(LRScheduler):
    method __init__ (line 29) | def __init__(
    method last_step (line 57) | def last_step(self):
    method last_step (line 61) | def last_step(self, value):
    method get_lr (line 64) | def get_lr(self):
  class InterpolatingLogScheduler (line 88) | class InterpolatingLogScheduler(LRScheduler):
    method __init__ (line 93) | def __init__(self, optimizer, num_steps, min_lr, max_lr, last_epoch=-1):
    method get_lr (line 113) | def get_lr(self):
  function _get_cosine_schedule_with_quadratic_warmup_lr_lambda (line 127) | def _get_cosine_schedule_with_quadratic_warmup_lr_lambda(
  function get_cosine_schedule_with_quadratic_warmup (line 144) | def get_cosine_schedule_with_quadratic_warmup(
  function _get_cosine_schedule_with_min_lr_lambda (line 182) | def _get_cosine_schedule_with_min_lr_lambda(
  function get_cosine_schedule_with_min_lr (line 201) | def get_cosine_schedule_with_min_lr(
  function _get_cosine_schedule_with_warmup_decay_constant_lr_lambda (line 222) | def _get_cosine_schedule_with_warmup_decay_constant_lr_lambda(
  function get_cosine_schedule_with_warmup_decay_constant (line 252) | def get_cosine_schedule_with_warmup_decay_constant(
  class JaggedLRRestartScheduler (line 299) | class JaggedLRRestartScheduler(LRScheduler):
    method __init__ (line 302) | def __init__(
    method get_lr (line 318) | def get_lr(self) -> float | Sequence[float]:

FILE: src/axolotl/utils/schemas/config.py
  class AxolotlInputConfig (line 57) | class AxolotlInputConfig(
    method datasets_serializer (line 1150) | def datasets_serializer(
    method warn_peft_trainable_token_to_fix_untrained (line 1159) | def warn_peft_trainable_token_to_fix_untrained(cls, data):
    method check_sageattn_wo_sample_packing (line 1179) | def check_sageattn_wo_sample_packing(cls, data):
    method check_sageattn_fft (line 1190) | def check_sageattn_fft(cls, data):
  class AxolotlConfigWCapabilities (line 1199) | class AxolotlConfigWCapabilities(AxolotlInputConfig):
    method check_bf16 (line 1206) | def check_bf16(self):
    method check_tf32 (line 1224) | def check_tf32(self):
    method check_fp8 (line 1230) | def check_fp8(self):
    method check_sample_packing_w_sdpa_bf16 (line 1241) | def check_sample_packing_w_sdpa_bf16(cls, data):
    method check_compute_capability_w_sageattn (line 1262) | def check_compute_capability_w_sageattn(cls, data):
    method check_multigpu_unsloth (line 1277) | def check_multigpu_unsloth(cls, data):
    method check_multigpu_lora_kernels (line 1292) | def check_multigpu_lora_kernels(cls, data):
    method check_quantize_moe_experts (line 1311) | def check_quantize_moe_experts(cls, data):
    method check_auto_enable_lora_kernels (line 1336) | def check_auto_enable_lora_kernels(cls, data):
    method check_adopt_torch_version (line 1392) | def check_adopt_torch_version(cls, data):
    method check_flex_torch_version (line 1410) | def check_flex_torch_version(cls, data):
    method check_torch_compile_auto (line 1428) | def check_torch_compile_auto(cls, data):
    method check_beta_and_trl_beta_match (line 1447) | def check_beta_and_trl_beta_match(cls, data):
    method check_min_torch_version (line 1454) | def check_min_torch_version(self):
    method check_qat_config (line 1466) | def check_qat_config(cls, data):
    method check_fsdp_torch_version (line 1495) | def check_fsdp_torch_version(cls, data):
    method default_dataloader_opts (line 1512) | def default_dataloader_opts(cls, data):
    method default_dataset_num_proc (line 1526) | def default_dataset_num_proc(cls, data):
    method check_deduplication_with_streaming (line 1546) | def check_deduplication_with_streaming(cls, data):
    method check_deduplication_with_skip_prepare (line 1557) | def check_deduplication_with_skip_prepare(cls, data):

FILE: src/axolotl/utils/schemas/datasets.py
  class UserDefinedPrompterType (line 11) | class UserDefinedPrompterType(BaseModel):
  class SFTDataset (line 39) | class SFTDataset(BaseModel):
    method handle_legacy_message_fields (line 200) | def handle_legacy_message_fields(cls, data):
    method check_chat_template_config (line 206) | def check_chat_template_config(cls, data):
  class PretrainingDataset (line 229) | class PretrainingDataset(BaseModel):
  class UserDefinedDPOType (line 242) | class UserDefinedDPOType(BaseModel):
  class DPODataset (line 254) | class DPODataset(BaseModel):
  class StepwiseSupervisedDataset (line 265) | class StepwiseSupervisedDataset(BaseModel):
  class UserDefinedKTOType (line 277) | class UserDefinedKTOType(BaseModel):
  class KTODataset (line 288) | class KTODataset(BaseModel):

FILE: src/axolotl/utils/schemas/deprecated.py
  class DeprecatedParameters (line 12) | class DeprecatedParameters(BaseModel):
    method validate_max_packed_sequence_len (line 27) | def validate_max_packed_sequence_len(cls, max_packed_sequence_len):
    method validate_rope_scaling (line 34) | def validate_rope_scaling(cls, rope_scaling):
    method validate_noisy_embedding_alpha (line 43) | def validate_noisy_embedding_alpha(cls, noisy_embedding_alpha):
    method validate_dpo_beta (line 50) | def validate_dpo_beta(cls, dpo_beta):
    method validate_evaluation_strategy (line 57) | def validate_evaluation_strategy(cls, evaluation_strategy):
    method validate_eval_table_size (line 64) | def validate_eval_table_size(cls, eval_table_size):
    method validate_eval_max_new_tokens (line 75) | def validate_eval_max_new_tokens(cls, eval_max_new_tokens):
    method validate_dpo_use_logits_to_keep (line 85) | def validate_dpo_use_logits_to_keep(cls, dpo_use_logits_to_keep):
    method validate_dpo_generate_during_eval (line 95) | def validate_dpo_generate_during_eval(cls, dpo_generate_during_eval):
  class RemappedParameters (line 104) | class RemappedParameters(BaseModel):

FILE: src/axolotl/utils/schemas/dynamic_checkpoint.py
  class DynamicCheckpointConfig (line 6) | class DynamicCheckpointConfig(BaseModel):

FILE: src/axolotl/utils/schemas/enums.py
  class TorchAOQuantDType (line 8) | class TorchAOQuantDType(Enum):
    method from_string (line 15) | def from_string(str):
  class RLType (line 28) | class RLType(str, Enum):
  class ChatTemplate (line 40) | class ChatTemplate(str, Enum):
  class CustomSupportedOptimizers (line 79) | class CustomSupportedOptimizers(str, Enum):
  class RingAttnFunc (line 97) | class RingAttnFunc(str, Enum):

FILE: src/axolotl/utils/schemas/fsdp.py
  class FSDPConfig (line 10) | class FSDPConfig(BaseModel):

FILE: src/axolotl/utils/schemas/integrations.py
  class MLFlowConfig (line 12) | class MLFlowConfig(BaseModel):
  class LISAConfig (line 33) | class LISAConfig(BaseModel):
  class WandbConfig (line 50) | class WandbConfig(BaseModel):
    method check_wandb_run (line 84) | def check_wandb_run(cls, data):
  class CometConfig (line 95) | class CometConfig(BaseModel):
  class GradioConfig (line 146) | class GradioConfig(BaseModel):
  class RayConfig (line 157) | class RayConfig(BaseModel):
  class OpenTelemetryConfig (line 181) | class OpenTelemetryConfig(BaseModel):
  class TrackioConfig (line 205) | class TrackioConfig(BaseModel):

FILE: src/axolotl/utils/schemas/internal/__init__.py
  class GPUCapabilities (line 8) | class GPUCapabilities(BaseModel):
  class EnvCapabilities (line 19) | class EnvCapabilities(BaseModel):

FILE: src/axolotl/utils/schemas/model.py
  class ModelInputConfig (line 12) | class ModelInputConfig(BaseModel):
    method hint_trust_remote_code (line 101) | def hint_trust_remote_code(cls, trust_remote_code):
  class ModelOutputConfig (line 109) | class ModelOutputConfig(BaseModel):
    method validate_save_safetensors (line 138) | def validate_save_safetensors(cls, v):
  class SpecialTokensConfig (line 149) | class SpecialTokensConfig(BaseModel):

FILE: src/axolotl/utils/schemas/multimodal.py
  class MultiModalConfig (line 9) | class MultiModalConfig(BaseModel):
    method convert_image_resize_algorithm (line 32) | def convert_image_resize_algorithm(cls, image_resize_algorithm):

FILE: src/axolotl/utils/schemas/peft.py
  class LoftQConfig (line 8) | class LoftQConfig(BaseModel):
  class PeftConfig (line 17) | class PeftConfig(BaseModel):
  class LoraConfig (line 28) | class LoraConfig(BaseModel):
    method validate_adapter (line 161) | def validate_adapter(cls, data):
    method validate_qlora (line 174) | def validate_qlora(self):
    method convert_loraplus_lr_embedding (line 200) | def convert_loraplus_lr_embedding(cls, loraplus_lr_embedding):
    method validate_lora_dropout (line 207) | def validate_lora_dropout(cls, data):
    method validate_lora_target_parameters_dropout (line 213) | def validate_lora_target_parameters_dropout(self):
  class ReLoRAConfig (line 226) | class ReLoRAConfig(BaseModel):

FILE: src/axolotl/utils/schemas/quantization.py
  function validate_ao_dtype (line 12) | def validate_ao_dtype(v: Any) -> TorchAOQuantDType | None:
  class QATConfig (line 31) | class QATConfig(BaseModel):
    method validate_dtype (line 57) | def validate_dtype(cls, v: Any) -> TorchAOQuantDType | None:
  class PTQConfig (line 61) | class PTQConfig(BaseModel):
    method validate_dtype (line 84) | def validate_dtype(cls, v: Any) -> TorchAOQuantDType | None:

FILE: src/axolotl/utils/schemas/training.py
  class LrGroup (line 15) | class LrGroup(BaseModel):
  class HyperparametersConfig (line 23) | class HyperparametersConfig(BaseModel):
    method hint_batch_size_set (line 168) | def hint_batch_size_set(cls, batch_size):
    method convert_learning_rate (line 179) | def convert_learning_rate(cls, learning_rate):
  class JaggedLRConfig (line 185) | class JaggedLRConfig(BaseModel):

FILE: src/axolotl/utils/schemas/trl.py
  class TRLConfig (line 8) | class TRLConfig(BaseModel):

FILE: src/axolotl/utils/schemas/utils.py
  function handle_legacy_message_fields_logic (line 8) | def handle_legacy_message_fields_logic(data: dict) -> dict:

FILE: src/axolotl/utils/schemas/validation.py
  class DatasetValidationMixin (line 22) | class DatasetValidationMixin:
    method set_default_seed (line 27) | def set_default_seed(cls, seed):
    method deprecate_sharegpt_datasets (line 35) | def deprecate_sharegpt_datasets(cls, datasets):
    method check_dataset_or_pretraining_dataset (line 57) | def check_dataset_or_pretraining_dataset(cls, data):
    method check_pretraining_streaming_deprecation (line 64) | def check_pretraining_streaming_deprecation(cls, data):
    method check_push_ds_auth (line 78) | def check_push_ds_auth(cls, data):
    method check_val_w_test_datasets (line 90) | def check_val_w_test_datasets(cls, data):
    method check_test_datasets_bench (line 99) | def check_test_datasets_bench(cls, data):
    method check_eval_packing (line 113) | def check_eval_packing(cls, data):
    method check_mm_prepare (line 148) | def check_mm_prepare(cls, data):
  class AttentionValidationMixin (line 159) | class AttentionValidationMixin:
    method check_attention_fields (line 164) | def check_attention_fields(cls, data):
    method check_sample_packing_without_attention (line 181) | def check_sample_packing_without_attention(cls, data):
    method check_sample_packing_with_s2attn (line 197) | def check_sample_packing_with_s2attn(cls, data):
    method check_scaling_softmax_requires_flex (line 207) | def check_scaling_softmax_requires_flex(cls, data):
  class TrainingValidationMixin (line 216) | class TrainingValidationMixin:
    method check_batch_size_fields (line 221) | def check_batch_size_fields(cls, data):
    method hint_sample_packing_padding (line 231) | def hint_sample_packing_padding(cls, data):
    method hint_reward_model_pad (line 247) | def hint_reward_model_pad(cls, data):
    method set_reward_model_defaults (line 258) | def set_reward_model_defaults(cls, data):
    method check_gas_bsz (line 275) | def check_gas_bsz(cls, data):
    method hint_eval_train_mbsz (line 284) | def hint_eval_train_mbsz(cls, data):
    method check_warmup (line 297) | def check_warmup(cls, data):
    method check_saves (line 304) | def check_saves(cls, data):
    method check_push_save (line 321) | def check_push_save(cls, data):
    method check_evals (line 332) | def check_evals(cls, data):
    method check_neftune (line 374) | def check_neftune(cls, data):
    method check_multipack_buffer_size (line 386) | def check_multipack_buffer_size(cls, data):
    method check_fft_possible_bad_config (line 409) | def check_fft_possible_bad_config(self):
    method check_fp8_config (line 427) | def check_fp8_config(cls, data):
    method check_use_reentrant_mismatch (line 457) | def check_use_reentrant_mismatch(cls, data):
    method check_eval_strategy (line 472) | def check_eval_strategy(cls, data):
    method check_causal_lm_evals (line 485) | def check_causal_lm_evals(cls, data):
    method check_tokenizer_use_mistral_common (line 503) | def check_tokenizer_use_mistral_common(cls, data):
    method check_mistral_common_import (line 522) | def check_mistral_common_import(cls, tokenizer_use_mistral_common):
    method check_mistral_common_incompatible_options (line 535) | def check_mistral_common_incompatible_options(cls, data):
    method pretrain_with_tps (line 565) | def pretrain_with_tps(cls, data):
  class LoRAValidationMixin (line 578) | class LoRAValidationMixin:
    method check_lr_groups (line 583) | def check_lr_groups(cls, data):
    method check_frozen (line 590) | def check_frozen(cls, data):
    method check_peft_layers_pattern (line 603) | def check_peft_layers_pattern(cls, data):
    method check_qlora_unsloth (line 612) | def check_qlora_unsloth(cls, data):
    method check_lora_axolotl_unsloth (line 626) | def check_lora_axolotl_unsloth(cls, data):
    method check_fused_lora (line 641) | def check_fused_lora(self):
    method warn_qlora_zero3_w_use_reentrant (line 648) | def warn_qlora_zero3_w_use_reentrant(cls, data):
    method check_lora_kernels_8bit (line 668) | def check_lora_kernels_8bit(cls, data):
    method check_lora_kernels_dora (line 683) | def check_lora_kernels_dora(cls, data):
    method check_lora_kernels_trust_remote_code (line 697) | def check_lora_kernels_trust_remote_code(cls, data):
  class RLValidationMixin (line 711) | class RLValidationMixin:
    method check_sample_packing_w_rl (line 716) | def check_sample_packing_w_rl(cls, data):
    method check_kto_config (line 723) | def check_kto_config(cls, data):
    method check_grpo_liger_sequence_parallel (line 734) | def check_grpo_liger_sequence_parallel(cls, data):
    method check_rl_config_gradient_checkpointing (line 746) | def check_rl_config_gradient_checkpointing(cls, data):
    method check_gdpo (line 770) | def check_gdpo(cls, data):
  class OptimizationValidationMixin (line 782) | class OptimizationValidationMixin:
    method check_adamw_optimizer_params (line 786) | def check_adamw_optimizer_params(self):
    method _resolve_fsdp_version (line 794) | def _resolve_fsdp_version(data):
    method check_muon_deepspeed_fsdp (line 803) | def check_muon_deepspeed_fsdp(cls, data):
    method check_flashoptim_deepspeed_fsdp (line 819) | def check_flashoptim_deepspeed_fsdp(cls, data):
    method check_batch_flattening_fa (line 838) | def check_batch_flattening_fa(cls, data):
    method check_xentropy_patch_conflicts (line 862) | def check_xentropy_patch_conflicts(cls, data):
    method check_cross_entropy_conflicts (line 873) | def check_cross_entropy_conflicts(cls, data):
    method check_fsdp_version (line 903) | def check_fsdp_version(cls, data):
    method check_fsdp2_cpu_offload_pin_memory (line 916) | def check_fsdp2_cpu_offload_pin_memory(cls, data):
    method check_fsdp2_base_model_quant_rl (line 933) | def check_fsdp2_base_model_quant_rl(cls, data):
    method check_fsdp_config_kwargs_prefix (line 949) | def check_fsdp_config_kwargs_prefix(cls, data):
    method check_fsdp_version_in_fsdp_config (line 971) | def check_fsdp_version_in_fsdp_config(cls, data):
    method check_fsdp_offload_w_8bit_optimizer (line 985) | def check_fsdp_offload_w_8bit_optimizer(self):
    method check_fsdp2_w_8bit_optimizer (line 1000) | def check_fsdp2_w_8bit_optimizer(self):
    method check_tensor_parallel_size_update_ds_json (line 1018) | def check_tensor_parallel_size_update_ds_json(cls, data):
    method check_deepcompile (line 1050) | def check_deepcompile(cls, data):
  class SystemValidationMixin (line 1069) | class SystemValidationMixin:
    method check_mem_mismatch (line 1074) | def check_mem_mismatch(cls, data):
    method check_fsdp_deepspeed (line 1086) | def check_fsdp_deepspeed(cls, data):
    method check_model_quantization_config_vs_bnb (line 1093) | def check_model_quantization_config_vs_bnb(cls, data):
    method check_npu_config (line 1103) | def check_npu_config(cls, data):
  class ChatTemplateValidationMixin (line 1136) | class ChatTemplateValidationMixin:
    method check_chat_template_config (line 1141) | def check_chat_template_config(cls, data):
  class PretrainingValidationMixin (line 1157) | class PretrainingValidationMixin:
    method check_pretraining_w_max_steps (line 1162) | def check_pretraining_w_max_steps(cls, data):
    method check_pretraining_w_group_by_length (line 1171) | def check_pretraining_w_group_by_length(cls, data):
    method check_pretraining_split_batches_accelerate (line 1180) | def check_pretraining_split_batches_accelerate(cls, data):
    method check_pretraining_w_val_set_size (line 1198) | def check_pretraining_w_val_set_size(cls, data):
    method check_streaming_w_val_set_size (line 1208) | def check_streaming_w_val_set_size(cls, data):
    method check_streaming_w_max_steps (line 1218) | def check_streaming_w_max_steps(cls, data):
    method check_streaming_w_multiple_datasets (line 1228) | def check_streaming_w_multiple_datasets(cls, data):
  class ModelCompatibilityValidationMixin (line 1241) | class ModelCompatibilityValidationMixin:
    method check_falcon_fsdp (line 1245) | def check_falcon_fsdp(self):
    method check_mpt_checkpointing (line 1251) | def check_mpt_checkpointing(self):
    method check_gradient_checkpointing_w_offload (line 1259) | def check_gradient_checkpointing_w_offload(self):
    method check_activation_offloading_wo_gc (line 1278) | def check_activation_offloading_wo_gc(self):
    method check_better_transformers (line 1284) | def check_better_transformers(self):
    method check_gptq_w_revision (line 1301) | def check_gptq_w_revision(cls, data):
    method check_gpt_oss_fsdp_loading (line 1312) | def check_gpt_oss_fsdp_loading(cls, data):
  class ComplexValidationMixin (line 1322) | class ComplexValidationMixin:
    method validate_neftune_noise_alpha (line 1327) | def validate_neftune_noise_alpha(cls, neftune_noise_alpha):
    method check_rl_beta (line 1333) | def check_rl_beta(self):
    method check_simpo_warmup (line 1340) | def check_simpo_warmup(self):
    method check_relora (line 1348) | def check_relora(self):
    method check_early_stopping (line 1371) | def check_early_stopping(self):
    method check_tensor_parallel_size (line 1384) | def check_tensor_parallel_size(self):
    method check_context_parallel_size (line 1390) | def check_context_parallel_size(self):
    method validate_ring_attn_func (line 1446) | def validate_ring_attn_func(self):
    method hint_gradient_checkpointing_dpo_lora_ddp (line 1463) | def hint_gradient_checkpointing_dpo_lora_ddp(self):
  class DistributedValidationMixin (line 1479) | class DistributedValidationMixin:
    method check_tensor_parallel_optimizer (line 1483) | def check_tensor_parallel_optimizer(self):
  class GRPOVllmValidationMixin (line 1493) | class GRPOVllmValidationMixin:
    method check_vllm_mode_set (line 1497) | def check_vllm_mode_set(self):
  class ValidationMixin (line 1506) | class ValidationMixin(

FILE: src/axolotl/utils/schemas/vllm.py
  class VllmConfig (line 8) | class VllmConfig(BaseModel):

FILE: src/axolotl/utils/tee.py
  class _FileOnlyWriter (line 23) | class _FileOnlyWriter(io.TextIOBase):
    method write (line 29) | def write(self, s: str) -> int:  # type: ignore[override]
    method flush (line 36) | def flush(self) -> None:  # type: ignore[override]
  class _StreamTee (line 48) | class _StreamTee(io.TextIOBase):
    method __init__ (line 54) | def __init__(self, stream: io.TextIOBase):
    method write (line 57) | def write(self, s: str) -> int:  # type: ignore[override]
    method flush (line 64) | def flush(self) -> None:  # type: ignore[override]
    method encoding (line 74) | def encoding(self):  # type: ignore[override]
    method errors (line 78) | def errors(self):  # type: ignore[override]
    method isatty (line 81) | def isatty(self):  # type: ignore[override]
    method fileno (line 84) | def fileno(self):  # type: ignore[override]
  function prepare_debug_log (line 90) | def prepare_debug_log(cfg, filename: str = "debug.log") -> str:
  function close_debug_log (line 140) | def close_debug_log() -> None:

FILE: src/axolotl/utils/tokenization.py
  function check_dataset_labels (line 10) | def check_dataset_labels(
  function check_example_labels (line 25) | def check_example_labels(example, tokenizer, text_only=False):
  function color_token_for_rl_debug (line 57) | def color_token_for_rl_debug(decoded_token, encoded_token, color, text_o...
  function process_tokens_for_rl_debug (line 67) | def process_tokens_for_rl_debug(tokens, color, tokenizer, text_only):
  function check_rl_example_labels (line 76) | def check_rl_example_labels(example, tokenizer, text_only=False):

FILE: src/axolotl/utils/trackio_.py
  function setup_trackio_env_vars (line 8) | def setup_trackio_env_vars(cfg: DictDefault):

FILE: src/axolotl/utils/train.py
  function determine_last_checkpoint (line 11) | def determine_last_checkpoint(cfg: DictDefault, update: bool = True) -> ...

FILE: src/axolotl/utils/trainer.py
  function weighted_cross_entropy (line 29) | def weighted_cross_entropy(
  function create_weighted_mask (line 47) | def create_weighted_mask(labels: torch.Tensor):
  function trainer_weighted_loss (line 78) | def trainer_weighted_loss(model_output, labels, shift_labels=True):
  function disable_datasets_caching (line 91) | def disable_datasets_caching():
  function add_position_ids (line 99) | def add_position_ids(sample):
  function add_pose_position_ids (line 138) | def add_pose_position_ids(
  function add_length (line 203) | def add_length(sample):
  function filter_sequences_by_length (line 208) | def filter_sequences_by_length(
  function process_datasets_for_packing (line 252) | def process_datasets_for_packing(cfg, train_dataset, eval_dataset):
  function process_pretraining_datasets_for_packing (line 386) | def process_pretraining_datasets_for_packing(
  function calculate_total_num_steps (line 408) | def calculate_total_num_steps(cfg, train_dataset, update=True):
  function setup_torch_compile_env (line 525) | def setup_torch_compile_env(cfg):
  function setup_deepspeed_env (line 533) | def setup_deepspeed_env(cfg, stage=None):
  function setup_fsdp_envs (line 589) | def setup_fsdp_envs(cfg):
  function setup_parallelism_envs (line 621) | def setup_parallelism_envs(cfg):
  function prepare_optim_env (line 643) | def prepare_optim_env(cfg):
  function setup_trainer (line 679) | def setup_trainer(

FILE: src/axolotl/utils/wandb_.py
  function setup_wandb_env_vars (line 8) | def setup_wandb_env_vars(cfg: DictDefault):

FILE: src/setuptools_axolotl_dynamic_dependencies.py
  function parse_requirements (line 12) | def parse_requirements():
  class BuildPyCommand (line 94) | class BuildPyCommand(_build_py):
    method finalize_options (line 99) | def finalize_options(self):

FILE: tests/cli/conftest.py
  function cli_runner (line 22) | def cli_runner():
  function valid_test_config (line 27) | def valid_test_config():
  function config_path (line 32) | def config_path(tmp_path):

FILE: tests/cli/test_cli_base.py
  class BaseCliTest (line 9) | class BaseCliTest:
    method _test_cli_validation (line 12) | def _test_cli_validation(self, cli_runner, command: str):
    method _test_basic_execution (line 30) | def _test_basic_execution(
    method _test_cli_overrides (line 78) | def _test_cli_overrides(self, tmp_path: Path, valid_test_config: str):

FILE: tests/cli/test_cli_evaluate.py
  class TestEvaluateCommand (line 10) | class TestEvaluateCommand(BaseCliTest):
    method test_evaluate_cli_validation (line 15) | def test_evaluate_cli_validation(self, cli_runner):
    method test_evaluate_basic_execution (line 19) | def test_evaluate_basic_execution(self, cli_runner, tmp_path, valid_te...
    method test_evaluate_basic_execution_no_accelerate (line 25) | def test_evaluate_basic_execution_no_accelerate(
    method test_evaluate_cli_overrides (line 47) | def test_evaluate_cli_overrides(self, cli_runner, tmp_path, valid_test...
    method test_evaluate_with_launcher_args_torchrun (line 73) | def test_evaluate_with_launcher_args_torchrun(
    method test_evaluate_with_launcher_args_accelerate (line 106) | def test_evaluate_with_launcher_args_accelerate(
    method test_evaluate_backward_compatibility_no_launcher_args (line 140) | def test_evaluate_backward_compatibility_no_launcher_args(

FILE: tests/cli/test_cli_fetch.py
  function test_fetch_cli_examples (line 8) | def test_fetch_cli_examples(cli_runner):
  function test_fetch_cli_deepspeed (line 17) | def test_fetch_cli_deepspeed(cli_runner):
  function test_fetch_cli_with_dest (line 26) | def test_fetch_cli_with_dest(cli_runner, tmp_path):
  function test_fetch_cli_invalid_directory (line 36) | def test_fetch_cli_invalid_directory(cli_runner):

FILE: tests/cli/test_cli_inference.py
  function test_inference_basic (line 8) | def test_inference_basic(cli_runner, config_path):
  function test_inference_gradio (line 21) | def test_inference_gradio(cli_runner, config_path):
  function test_inference_with_launcher_args_torchrun (line 34) | def test_inference_with_launcher_args_torchrun(cli_runner, config_path):
  function test_inference_with_launcher_args_accelerate (line 63) | def test_inference_with_launcher_args_accelerate(cli_runner, config_path):
  function test_inference_gradio_with_launcher_args (line 93) | def test_inference_gradio_with_launcher_args(cli_runner, config_path):
  function test_inference_backward_compatibility_no_launcher_args (line 123) | def test_inference_backward_compatibility_no_launcher_args(cli_runner, c...

FILE: tests/cli/test_cli_interface.py
  function test_build_command (line 6) | def test_build_command():
  function test_invalid_command_options (line 28) | def test_invalid_command_options(cli_runner):
  function test_required_config_argument (line 43) | def test_required_config_argument(cli_runner):

FILE: tests/cli/test_cli_merge_lora.py
  function test_merge_lora_basic (line 8) | def test_merge_lora_basic(cli_runner, config_path):
  function test_merge_lora_with_dirs (line 18) | def test_merge_lora_with_dirs(cli_runner, config_path, tmp_path):
  function test_merge_lora_nonexistent_config (line 44) | def test_merge_lora_nonexistent_config(cli_runner, tmp_path):
  function test_merge_lora_nonexistent_lora_dir (line 51) | def test_merge_lora_nonexistent_lora_dir(cli_runner, config_path, tmp_pa...

FILE: tests/cli/test_cli_merge_sharded_fsdp_weights.py
  function test_merge_sharded_fsdp_weights_no_accelerate (line 8) | def test_merge_sharded_fsdp_weights_no_accelerate(cli_runner, config_path):
  function test_merge_sharded_fsdp_weights_with_launcher_args_torchrun (line 21) | def test_merge_sharded_fsdp_weights_with_launcher_args_torchrun(
  function test_merge_sharded_fsdp_weights_with_launcher_args_accelerate (line 52) | def test_merge_sharded_fsdp_weights_with_launcher_args_accelerate(
  function test_merge_sharded_fsdp_weights_backward_compatibility_no_launcher_args (line 84) | def test_merge_sharded_fsdp_weights_backward_compatibility_no_launcher_a...

FILE: tests/cli/test_cli_preprocess.py
  function cleanup_last_run_prepared (line 13) | def cleanup_last_run_prepared():
  function test_preprocess_config_not_found (line 20) | def test_preprocess_config_not_found(cli_runner):
  function test_preprocess_basic (line 26) | def test_preprocess_basic(cli_runner, config_path):
  function test_preprocess_without_download (line 40) | def test_preprocess_without_download(cli_runner, config_path):
  function test_preprocess_custom_path (line 53) | def test_preprocess_custom_path(cli_runner, tmp_path, valid_test_config):

FILE: tests/cli/test_cli_sweeps.py
  function test_generate_sweep_configs_no_pairs (line 8) | def test_generate_sweep_configs_no_pairs():
  function test_generate_sweep_configs_with_pairs (line 33) | def test_generate_sweep_configs_with_pairs():

FILE: tests/cli/test_cli_train.py
  class TestTrainCommand (line 10) | class TestTrainCommand(BaseCliTest):
    method test_train_cli_validation (line 15) | def test_train_cli_validation(self, cli_runner):
    method test_train_basic_execution (line 19) | def test_train_basic_execution(self, cli_runner, tmp_path, valid_test_...
    method test_train_basic_execution_no_accelerate (line 25) | def test_train_basic_execution_no_accelerate(
    method test_train_cli_overrides (line 51) | def test_train_cli_overrides(self, cli_runner, tmp_path, valid_test_co...
    method test_train_with_launcher_args_torchrun (line 79) | def test_train_with_launcher_args_torchrun(
    method test_train_with_launcher_args_accelerate (line 112) | def test_train_with_launcher_args_accelerate(
    method test_train_backward_compatibility_no_launcher_args (line 147) | def test_train_backward_compatibility_no_launcher_args(
    method test_train_mixed_args_with_launcher_args (line 182) | def test

[
  {
    "path": ".axolotl-complete.bash",
    "chars": 1435,
    "preview": "#!/bin/bash\n\n_axolotl_completions() {\n    local cur prev\n    COMPREPLY=()\n    cur=\"${COMP_WORDS[COMP_CWORD]}\"\n    prev=\""
  },
  {
    "path": ".bandit",
    "chars": 53,
    "preview": "[bandit]\nexclude = tests\nskips = B101,B615,B102,B110\n"
  },
  {
    "path": ".coderabbit.yaml",
    "chars": 412,
    "preview": "# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json\nlanguage: \"en-US\"\nearly_access: false\n"
  },
  {
    "path": ".coveragerc",
    "chars": 213,
    "preview": "[run]\nsource = axolotl\nomit =\n    */tests/*\n    setup.py\n\n[report]\nexclude_lines =\n    pragma: no cover\n    def __repr__"
  },
  {
    "path": ".editorconfig",
    "chars": 186,
    "preview": "root = true\n\n[*]\nend_of_line = lf\ninsert_final_newline = true\ntrim_trailing_whitespace = true\n\n[*.py]\nindent_style = spa"
  },
  {
    "path": ".gitattributes",
    "chars": 49,
    "preview": "data/*.jsonl filter=lfs diff=lfs merge=lfs -text\n"
  },
  {
    "path": ".github/CODE_OF_CONDUCT.md",
    "chars": 5242,
    "preview": "# Contributor Covenant Code of Conduct\n\n## Our Pledge\n\nWe as members, contributors, and leaders pledge to make participa"
  },
  {
    "path": ".github/CONTRIBUTING.md",
    "chars": 4073,
    "preview": "# Contributing to axolotl\n\nFirst of all, thank you for your interest in contributing to axolotl! We appreciate the time "
  },
  {
    "path": ".github/FUNDING.yml",
    "chars": 803,
    "preview": "# These are supported funding model platforms\n\ngithub: # Replace with up to 4 GitHub Sponsors-enabled usernames e.g., [u"
  },
  {
    "path": ".github/ISSUE_TEMPLATE/bug-report.yaml",
    "chars": 3509,
    "preview": "name: Bug Report\ndescription: File a bug report\nlabels: [\"bug\", \"needs triage\"]\nbody:\n  - type: markdown\n    attributes:"
  },
  {
    "path": ".github/ISSUE_TEMPLATE/config.yml",
    "chars": 294,
    "preview": "blank_issues_enabled: false\ncontact_links:\n  - name: Ask a question\n    url: https://github.com/axolotl-ai-cloud/axolotl"
  },
  {
    "path": ".github/ISSUE_TEMPLATE/docs.yml",
    "chars": 1873,
    "preview": "name: Documentation Improvement / Clarity\ndescription: Make a suggestion to improve the project documentation.\nlabels: ["
  },
  {
    "path": ".github/ISSUE_TEMPLATE/feature-request.yaml",
    "chars": 2466,
    "preview": "name: Feature Request / Enhancement\ndescription: Suggest a new feature or feature enhancement for the project\nlabels: [\""
  },
  {
    "path": ".github/PULL_REQUEST_TEMPLATE.md",
    "chars": 1049,
    "preview": "<!--- Provide a general summary of your changes in the Title above -->\n\n# Description\n\n<!--- Describe your changes in de"
  },
  {
    "path": ".github/SECURITY.md",
    "chars": 407,
    "preview": "# Security Policy\n\n## Supported Versions\n\nDue to the nature of the fast development that is happening in this project, o"
  },
  {
    "path": ".github/SUPPORT.md",
    "chars": 474,
    "preview": "# Support\n\nIf you need help with this project or have questions, please:\n\n1. Check the documentation.\n2. Search the exis"
  },
  {
    "path": ".github/release-drafter.yml",
    "chars": 632,
    "preview": "name-template: 'v$RESOLVED_VERSION'\ntag-template: 'v$RESOLVED_VERSION'\ncategories:\n  - title: '🚀 Features'\n    labels:\n "
  },
  {
    "path": ".github/workflows/base.yml",
    "chars": 10420,
    "preview": "name: ci-cd-base\n\non:\n  push:\n    branches:\n      - \"main\"\n    paths:\n      - 'docker/Dockerfile-base'\n      - 'docker/D"
  },
  {
    "path": ".github/workflows/docs.yml",
    "chars": 1030,
    "preview": "name: Publish Docs\non:\n  push:\n    branches:\n      - main\n\npermissions:\n    contents: write\n    pages: write\n\njobs:\n    "
  },
  {
    "path": ".github/workflows/lint.yml",
    "chars": 715,
    "preview": "name: lint\non:\n  # check on PRs, and manual triggers\n  merge_group:\n  pull_request:\n      types: [opened, synchronize, r"
  },
  {
    "path": ".github/workflows/main.yml",
    "chars": 14905,
    "preview": "name: ci-cd\n\non:\n  push:\n    branches:\n      - \"main\"\n    tags:\n      - \"v*\"\n  workflow_dispatch:\n\npermissions:\n  conten"
  },
  {
    "path": ".github/workflows/multi-gpu-e2e.yml",
    "chars": 3018,
    "preview": "name: docker-multigpu-tests-biweekly\n\non:\n  pull_request:\n    paths:\n      - 'tests/e2e/multigpu/**.py'\n      - 'require"
  },
  {
    "path": ".github/workflows/nightlies.yml",
    "chars": 4037,
    "preview": "name: docker-nightlies\n\non:\n  workflow_dispatch:\n  schedule:\n    - cron: '0 0 * * *'  # Runs at 00:00 UTC every day\n\nper"
  },
  {
    "path": ".github/workflows/precommit-autoupdate.yml",
    "chars": 1113,
    "preview": "name: Pre-commit auto-update\n\non:\n  schedule:\n    - cron: '0 0 1 * *'  # Run monthly\n  workflow_dispatch:  # Manual kick"
  },
  {
    "path": ".github/workflows/preview-docs.yml",
    "chars": 2377,
    "preview": "name: Preview\non:\n  workflow_dispatch:\n  pull_request:\n    types: [opened, synchronize, reopened, ready_for_review]\n\n   "
  },
  {
    "path": ".github/workflows/pypi.yml",
    "chars": 1617,
    "preview": "name: publish pypi\n\non:\n  push:\n    tags:\n      - \"v*\"\n  workflow_dispatch:\n\npermissions: {}\n\njobs:\n  setup_release:\n   "
  },
  {
    "path": ".github/workflows/tests-nightly.yml",
    "chars": 7380,
    "preview": "name: Tests Nightly against upstream main\non:\n  workflow_dispatch:\n  schedule:\n    - cron: '0 0 * * *'  # Runs at 00:00 "
  },
  {
    "path": ".github/workflows/tests.yml",
    "chars": 14804,
    "preview": "name: Tests\non:\n  # check on push/merge to main, PRs, and manual triggers\n  merge_group:\n  push:\n    branches:\n      - \""
  },
  {
    "path": ".gitignore",
    "chars": 3445,
    "preview": "**/axolotl.egg-info\nconfigs\nlast_run_prepared/\noutputs\n.vscode\n_site/\n\n# Byte-compiled / optimized / DLL files\n__pycache"
  },
  {
    "path": ".mypy.ini",
    "chars": 899,
    "preview": "[mypy]\nplugins = pydantic.mypy\nexclude = venv\n\n[mypy-alpaca_lora_4bit.*]\nignore_missing_imports = True\n\n[mypy-axolotl.mo"
  },
  {
    "path": ".pre-commit-config.yaml",
    "chars": 799,
    "preview": "default_language_version:\n    python: python3\n\nrepos:\n-   repo: https://github.com/pre-commit/pre-commit-hooks\n    rev: "
  },
  {
    "path": ".runpod/.gitignore",
    "chars": 3102,
    "preview": "# Byte-compiled / optimized / DLL files\n__pycache__/\n*.py[cod]\n*$py.class\n\n# C extensions\n*.so\n\n# Distribution / packagi"
  },
  {
    "path": ".runpod/Dockerfile",
    "chars": 662,
    "preview": "FROM axolotlai/axolotl-cloud:main-py3.11-cu124-2.6.0\n\nCOPY .runpod/requirements.txt /requirements.txt\nRUN --mount=type=c"
  },
  {
    "path": ".runpod/README.md",
    "chars": 14939,
    "preview": "<h1>LLM Post Training- Full fine-tune, LoRA, QLoRa etc. Llama/Mistral/Gemma and more</h1>\n\n# Configuration Options\n\nThis"
  },
  {
    "path": ".runpod/hub.json",
    "chars": 2334,
    "preview": "{\n  \"title\": \"Axolotl Fine-Tuning\",\n  \"description\": \"Serverless fine-tuning of open-source LLMs with Axolotl. Supports "
  },
  {
    "path": ".runpod/requirements.txt",
    "chars": 341,
    "preview": "# Required Python packages get listed here, one per line.\n# Reccomended to lock the version number to avoid unexpected c"
  },
  {
    "path": ".runpod/src/config/config.yaml",
    "chars": 20840,
    "preview": "# # This is the huggingface model that contains *.pt, *.safetensors, or *.bin files\n# # This can also be a relative path"
  },
  {
    "path": ".runpod/src/handler.py",
    "chars": 1865,
    "preview": "\"\"\"\nRunpod serverless entrypoint handler\n\"\"\"\n\nimport os\n\nimport runpod\nimport yaml\nfrom huggingface_hub._login import lo"
  },
  {
    "path": ".runpod/src/test_input.json",
    "chars": 1557,
    "preview": "{\n  \"input\": {\n    \"user_id\": \"user\",\n    \"model_id\": \"llama-test\",\n    \"run_id\": \"llama-test\",\n    \"credentials\": {\n   "
  },
  {
    "path": ".runpod/src/train.py",
    "chars": 1430,
    "preview": "\"\"\"\nRunpod train entrypoint\n\"\"\"\n\nimport asyncio\n\n\nasync def train(config_path: str, gpu_id: str = \"0\", preprocess: bool "
  },
  {
    "path": ".runpod/src/utils.py",
    "chars": 2740,
    "preview": "\"\"\"\nRunpod launcher utils\n\"\"\"\n\nimport os\n\nimport yaml\n\n\ndef get_output_dir(run_id):\n    path = f\"fine-tuning/{run_id}\"\n "
  },
  {
    "path": ".runpod/test-input.json",
    "chars": 2024,
    "preview": "{\n  \"input\": {\n    \"name\": \"quick_smoke_test_sft\",\n    \"user_id\": \"user\",\n    \"model_id\": \"llama-test\",\n    \"run_id\": \"l"
  },
  {
    "path": ".runpod/tests.json",
    "chars": 2297,
    "preview": "{\n  \"tests\": [\n    {\n      \"name\": \"quick_smoke_test_sft\",\n      \"input\": {\n        \"user_id\": \"user\",\n        \"model_id"
  },
  {
    "path": "CITATION.cff",
    "chars": 340,
    "preview": "cff-version: 1.2.0\ntype: software\ntitle: \"Axolotl: Open Source LLM Post-Training\"\nmessage: \"If you use this software, pl"
  },
  {
    "path": "CNAME",
    "chars": 16,
    "preview": "docs.axolotl.ai\n"
  },
  {
    "path": "FAQS.md",
    "chars": 648,
    "preview": "# FAQs\n\n- Can you train StableLM with this? Yes, but only with a single GPU atm. Multi GPU support is coming soon! Just "
  },
  {
    "path": "LICENSE",
    "chars": 11358,
    "preview": "\n                                 Apache License\n                           Version 2.0, January 2004\n                  "
  },
  {
    "path": "MANIFEST.in",
    "chars": 204,
    "preview": "include requirements.txt\ninclude README.md\ninclude LICENSE\ninclude src/setuptools_axolotl_dynamic_dependencies.py\ninclud"
  },
  {
    "path": "README.md",
    "chars": 15284,
    "preview": "<p align=\"center\">\n    <picture>\n        <source media=\"(prefers-color-scheme: dark)\" srcset=\"https://raw.githubusercont"
  },
  {
    "path": "VERSION",
    "chars": 12,
    "preview": "0.16.0.dev0\n"
  },
  {
    "path": "_quarto.yml",
    "chars": 11132,
    "preview": "project:\n  type: website\n  pre-render:\n   - docs/scripts/generate_config_docs.py\n   - docs/scripts/generate_examples_doc"
  },
  {
    "path": "benchmarks/bench_entropy.py",
    "chars": 6320,
    "preview": "\"\"\"Benchmark for entropy_from_logits Triton kernel vs original chunked implementation.\n\nUsage: CUDA_VISIBLE_DEVICES=0 py"
  },
  {
    "path": "benchmarks/bench_scattermoe_lora.py",
    "chars": 8845,
    "preview": "\"\"\"Benchmark for ScatterMoE LoRA Triton kernels.\n\nMeasures forward, backward dX, and backward dA/dB kernels at common Mo"
  },
  {
    "path": "benchmarks/bench_selective_logsoftmax.py",
    "chars": 5661,
    "preview": "\"\"\"Benchmark for selective_log_softmax Triton kernel vs original implementation.\n\nUsage: CUDA_VISIBLE_DEVICES=0 python b"
  },
  {
    "path": "cicd/Dockerfile-uv.jinja",
    "chars": 2355,
    "preview": "FROM axolotlai/axolotl-base-uv:{{ BASE_TAG }}\n\nENV TORCH_CUDA_ARCH_LIST=\"7.0 7.5 8.0 8.6 9.0+PTX\"\nENV AXOLOTL_EXTRAS=\"{{"
  },
  {
    "path": "cicd/Dockerfile.jinja",
    "chars": 2347,
    "preview": "FROM axolotlai/axolotl-base:{{ BASE_TAG }}\n\nENV TORCH_CUDA_ARCH_LIST=\"7.0 7.5 8.0 8.6 8.7 8.9 9.0+PTX\"\nENV AXOLOTL_EXTRA"
  },
  {
    "path": "cicd/__init__.py",
    "chars": 0,
    "preview": ""
  },
  {
    "path": "cicd/cicd.sh",
    "chars": 1901,
    "preview": "#!/bin/bash\nset -e\n\npython -c \"import torch; assert '$PYTORCH_VERSION' in torch.__version__\"\n\ncurl -L https://axolotl-ci"
  },
  {
    "path": "cicd/cleanup.py",
    "chars": 358,
    "preview": "\"\"\"Modal app to run axolotl GPU cleanup\"\"\"\n\nfrom .single_gpu import VOLUME_CONFIG, app, cicd_image, run_cmd\n\n\n@app.funct"
  },
  {
    "path": "cicd/cleanup.sh",
    "chars": 297,
    "preview": "#!/bin/bash\nset -e\n\n# cleanup old cache files for datasets processing and intermediate mappings\nfind /workspace/data/hug"
  },
  {
    "path": "cicd/e2e_tests.py",
    "chars": 404,
    "preview": "\"\"\"Modal app to run axolotl GPU tests\"\"\"\n\nfrom .single_gpu import GPU_CONFIG, VOLUME_CONFIG, app, cicd_image, run_cmd\n\n\n"
  },
  {
    "path": "cicd/multigpu.py",
    "chars": 2403,
    "preview": "\"\"\"\nmodal application to run axolotl gpu tests in Modal\n\"\"\"\n\nimport os\nimport pathlib\nimport tempfile\n\nimport jinja2\nimp"
  },
  {
    "path": "cicd/multigpu.sh",
    "chars": 852,
    "preview": "#!/bin/bash\nset -e\n\n# Only run two tests at a time to avoid OOM on GPU (with coverage collection)\npytest -v --durations="
  },
  {
    "path": "cicd/single_gpu.py",
    "chars": 2370,
    "preview": "\"\"\"Modal app to run axolotl GPU tests\"\"\"\n\nimport os\nimport pathlib\nimport tempfile\n\nimport jinja2\nimport modal\nimport mo"
  },
  {
    "path": "codecov.yml",
    "chars": 1046,
    "preview": "codecov:\n  require_ci_to_pass: yes\n  notify:\n    wait_for_ci: true\n\ncoverage:\n  precision: 2\n  round: down\n  range: \"70."
  },
  {
    "path": "deepspeed_configs/zero1.json",
    "chars": 484,
    "preview": "{\n  \"zero_optimization\": {\n    \"stage\": 1,\n    \"overlap_comm\": true\n  },\n  \"bf16\": {\n    \"enabled\": \"auto\"\n  },\n  \"fp16\""
  },
  {
    "path": "deepspeed_configs/zero1_torch_compile.json",
    "chars": 552,
    "preview": "{\n  \"zero_optimization\": {\n    \"stage\": 1,\n    \"overlap_comm\": true\n  },\n  \"bf16\": {\n    \"enabled\": \"auto\"\n  },\n  \"fp16\""
  },
  {
    "path": "deepspeed_configs/zero2.json",
    "chars": 574,
    "preview": "{\n  \"zero_optimization\": {\n    \"stage\": 2,\n    \"offload_optimizer\": {\n      \"device\": \"cpu\"\n    },\n    \"contiguous_gradi"
  },
  {
    "path": "deepspeed_configs/zero2_torch_compile.json",
    "chars": 642,
    "preview": "{\n  \"compile\": {\n    \"disable\": false,\n    \"backend\": \"inductor\"\n  },\n  \"zero_optimization\": {\n    \"stage\": 2,\n    \"offl"
  },
  {
    "path": "deepspeed_configs/zero3.json",
    "chars": 777,
    "preview": "{\n  \"zero_optimization\": {\n    \"stage\": 3,\n    \"overlap_comm\": true,\n    \"contiguous_gradients\": true,\n    \"sub_group_si"
  },
  {
    "path": "deepspeed_configs/zero3_bf16.json",
    "chars": 583,
    "preview": "{\n  \"zero_optimization\": {\n    \"stage\": 3,\n    \"overlap_comm\": true,\n    \"contiguous_gradients\": true,\n    \"sub_group_si"
  },
  {
    "path": "deepspeed_configs/zero3_bf16_cpuoffload_all.json",
    "chars": 824,
    "preview": "{\n  \"zero_force_ds_cpu_optimizer\": false,\n  \"zero_allow_untested_optimizer\": true,\n  \"zero_optimization\": {\n    \"stage\":"
  },
  {
    "path": "deepspeed_configs/zero3_bf16_cpuoffload_params.json",
    "chars": 742,
    "preview": "{\n  \"zero_force_ds_cpu_optimizer\": false,\n  \"zero_allow_untested_optimizer\": true,\n  \"zero_optimization\": {\n    \"stage\":"
  },
  {
    "path": "devtools/README.md",
    "chars": 158,
    "preview": "This directory contains example config files that might be useful for debugging. Please see [docs/debugging.qmd](../docs"
  },
  {
    "path": "devtools/dev_chat_template.yml",
    "chars": 939,
    "preview": "# Example config for debugging the chat_template prompt format\nbase_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0\nmodel_type"
  },
  {
    "path": "docker/Dockerfile",
    "chars": 1691,
    "preview": "ARG BASE_TAG=main-base\nFROM axolotlai/axolotl-base:$BASE_TAG\n\nARG TORCH_CUDA_ARCH_LIST=\"7.0 7.5 8.0 8.6+PTX\"\nARG AXOLOTL"
  },
  {
    "path": "docker/Dockerfile-base",
    "chars": 3340,
    "preview": "ARG CUDA_VERSION=\"11.8.0\"\nARG CUDNN_VERSION=\"8\"\nARG UBUNTU_VERSION=\"22.04\"\nARG MAX_JOBS=4\nARG TARGETARCH\n\nFROM nvidia/cu"
  },
  {
    "path": "docker/Dockerfile-base-next",
    "chars": 1501,
    "preview": "ARG CUDA_VERSION=\"12.8.1\"\nARG CUDNN_VERSION=\"8\"\nARG UBUNTU_VERSION=\"22.04\"\nARG MAX_JOBS=4\n\nFROM nvidia/cuda:$CUDA_VERSIO"
  },
  {
    "path": "docker/Dockerfile-base-nightly",
    "chars": 1849,
    "preview": "ARG CUDA_VERSION=\"12.8.1\"\nARG CUDNN_VERSION=\"8\"\nARG UBUNTU_VERSION=\"22.04\"\nARG MAX_JOBS=4\n\nFROM nvidia/cuda:$CUDA_VERSIO"
  },
  {
    "path": "docker/Dockerfile-cloud",
    "chars": 1139,
    "preview": "ARG BASE_TAG=main\nFROM axolotlai/axolotl:$BASE_TAG\n\nENV HF_DATASETS_CACHE=\"/workspace/data/huggingface-cache/datasets\"\nE"
  },
  {
    "path": "docker/Dockerfile-cloud-no-tmux",
    "chars": 1037,
    "preview": "ARG BASE_TAG=main\nFROM axolotlai/axolotl:$BASE_TAG\n\nENV HF_DATASETS_CACHE=\"/workspace/data/huggingface-cache/datasets\"\nE"
  },
  {
    "path": "docker/Dockerfile-cloud-uv",
    "chars": 1145,
    "preview": "ARG BASE_TAG=main\nFROM axolotlai/axolotl-uv:$BASE_TAG\n\nENV HF_DATASETS_CACHE=\"/workspace/data/huggingface-cache/datasets"
  },
  {
    "path": "docker/Dockerfile-tests",
    "chars": 1183,
    "preview": "ARG BASE_TAG=main-base\nFROM axolotlai/axolotl-base:$BASE_TAG\n\nARG TORCH_CUDA_ARCH_LIST=\"7.0 7.5 8.0 8.6+PTX\"\nARG AXOLOTL"
  },
  {
    "path": "docker/Dockerfile-uv",
    "chars": 1713,
    "preview": "ARG BASE_TAG=main-base\nFROM axolotlai/axolotl-base-uv:$BASE_TAG\n\nARG TORCH_CUDA_ARCH_LIST=\"7.0 7.5 8.0 8.6+PTX\"\nARG AXOL"
  },
  {
    "path": "docker/Dockerfile-uv-base",
    "chars": 2113,
    "preview": "ARG CUDA_VERSION=\"12.6.3\"\nARG CUDNN_VERSION=\"\"\nARG UBUNTU_VERSION=\"22.04\"\nARG MAX_JOBS=4\nARG TARGETARCH\n\nFROM nvidia/cud"
  },
  {
    "path": "docker-compose.yaml",
    "chars": 701,
    "preview": "# version: '3.8'\nservices:\n  axolotl:\n    build:\n      context: .\n      dockerfile: ./docker/Dockerfile\n    volumes:\n   "
  },
  {
    "path": "docs/.gitignore",
    "chars": 94,
    "preview": "/.quarto/\n_site/\n/api/*.qmd\n/api/*.html\nconfig-reference.qmd\nmodels/**/*.qmd\nmodels/**/*.html\n"
  },
  {
    "path": "docs/amd_hpc.qmd",
    "chars": 3341,
    "preview": "---\ntitle: AMD GPUs on HPC Systems\ndescription: A comprehensive guide for using Axolotl on distributed systems with AMD "
  },
  {
    "path": "docs/attention.qmd",
    "chars": 3820,
    "preview": "---\ntitle: Attention\ndescription: Supported attention modules in Axolotl\n---\n\n## SDP Attention\n\nThis is the default buil"
  },
  {
    "path": "docs/batch_vs_grad.qmd",
    "chars": 3008,
    "preview": "---\ntitle: Batch size vs Gradient accumulation\ndescription: Understanding of batch size and gradient accumulation steps\n"
  },
  {
    "path": "docs/checkpoint_saving.qmd",
    "chars": 2394,
    "preview": "---\ntitle: \"Checkpoint Saving\"\nformat:\n  html:\n    toc: true\n    toc-depth: 2\n    number-sections: true\nexecute:\n  enabl"
  },
  {
    "path": "docs/cli.qmd",
    "chars": 8582,
    "preview": "---\ntitle: \"Command Line Interface (CLI)\"\nformat:\n  html:\n    toc: true\n    toc-expand: 2\n    toc-depth: 3\nexecute:\n  en"
  },
  {
    "path": "docs/custom_integrations.qmd",
    "chars": 3984,
    "preview": "---\ntitle: Custom Integrations\ntoc: true\ntoc-depth: 3\n---\n\n```{python}\n#| echo: false\n\nimport os\nimport re\n\ndef process_"
  },
  {
    "path": "docs/dataset-formats/conversation.qmd",
    "chars": 9552,
    "preview": "---\ntitle: Conversation\ndescription: Conversation format for supervised fine-tuning.\norder: 3\n---\n\n## chat_template\n\nCha"
  },
  {
    "path": "docs/dataset-formats/index.qmd",
    "chars": 17703,
    "preview": "---\ntitle: Dataset Formats\ndescription: Guide to Dataset Formats in Axolotl\nback-to-top-navigation: true\ntoc: true\ntoc-d"
  },
  {
    "path": "docs/dataset-formats/inst_tune.qmd",
    "chars": 3818,
    "preview": "---\ntitle: Instruction Tuning\ndescription: Instruction tuning formats for supervised fine-tuning.\norder: 2\n---\n\n## alpac"
  },
  {
    "path": "docs/dataset-formats/pretraining.qmd",
    "chars": 764,
    "preview": "---\ntitle: Pre-training\ndescription: Data format for a pre-training completion task.\norder: 1\n---\n\nFor pretraining, ther"
  },
  {
    "path": "docs/dataset-formats/stepwise_supervised.qmd",
    "chars": 683,
    "preview": "---\ntitle: Stepwise Supervised Format\ndescription: Format for datasets with stepwise completions and labels\norder: 3\n---"
  },
  {
    "path": "docs/dataset-formats/template_free.qmd",
    "chars": 6465,
    "preview": "---\ntitle: Template-Free\ndescription: Construct prompts without a template.\ntoc: true\ntoc-depth: 3\norder: 4\n---\n\n## Back"
  },
  {
    "path": "docs/dataset-formats/tokenized.qmd",
    "chars": 910,
    "preview": "---\ntitle: Custom Pre-Tokenized Dataset\ndescription: How to use a custom pre-tokenized dataset.\norder: 5\n---\n\n- Pass an "
  },
  {
    "path": "docs/dataset_loading.qmd",
    "chars": 6700,
    "preview": "---\ntitle: Dataset Loading\ndescription: Understanding how to load datasets from different sources\nback-to-top-navigation"
  },
  {
    "path": "docs/dataset_preprocessing.qmd",
    "chars": 2025,
    "preview": "---\ntitle: Dataset Preprocessing\ndescription: How datasets are processed\n---\n\n## Overview\n\nDataset pre-processing is the"
  },
  {
    "path": "docs/debugging.qmd",
    "chars": 14344,
    "preview": "---\ntitle: Debugging\ndescription: How to debug Axolotl\n---\n\n\nThis document provides some tips and tricks for debugging A"
  },
  {
    "path": "docs/docker.qmd",
    "chars": 3189,
    "preview": "---\ntitle: \"Docker\"\nformat:\n  html:\n    toc: true\n    toc-depth: 4\n---\n\nThis section describes the different Docker imag"
  },
  {
    "path": "docs/expert_quantization.qmd",
    "chars": 2831,
    "preview": "---\ntitle: \"MoE Expert Quantization\"\ndescription: \"Reduce VRAM usage when training MoE model adapters by quantizing expe"
  },
  {
    "path": "docs/faq.qmd",
    "chars": 8564,
    "preview": "---\ntitle: FAQ\ndescription: Frequently asked questions\n---\n\n### General\n\n**Q: The trainer stopped and hasn't progressed "
  },
  {
    "path": "docs/fsdp_qlora.qmd",
    "chars": 2221,
    "preview": "---\ntitle: \"FSDP + QLoRA\"\ndescription: Use FSDP with QLoRA to fine-tune large LLMs on consumer GPUs.\nformat:\n  html:\n   "
  },
  {
    "path": "docs/getting-started.qmd",
    "chars": 4624,
    "preview": "---\ntitle: \"Quickstart\"\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n    number-sections: true\nexecute:\n  enabled: fal"
  },
  {
    "path": "docs/gradient_checkpointing.qmd",
    "chars": 997,
    "preview": "---\ntitle: Gradient Checkpointing and Activation Offloading\n---\n\nGradient checkpointing and activation offloading are te"
  },
  {
    "path": "docs/inference.qmd",
    "chars": 2824,
    "preview": "---\ntitle: \"Inference and Merging\"\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n    number-sections: true\nexecute:\n  e"
  },
  {
    "path": "docs/input_output.qmd",
    "chars": 200,
    "preview": "---\ntitle: Template-free prompt construction\ndescription: \"Template-free prompt construction with the `input_output` for"
  },
  {
    "path": "docs/installation.qmd",
    "chars": 5355,
    "preview": "---\ntitle: \"Installation\"\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n    number-sections: true\nexecute:\n  enabled: f"
  },
  {
    "path": "docs/lora_optims.qmd",
    "chars": 4694,
    "preview": "---\ntitle: \"LoRA Optimizations\"\ndescription: \"Custom autograd functions and Triton kernels in Axolotl for optimized LoRA"
  },
  {
    "path": "docs/lr_groups.qmd",
    "chars": 1041,
    "preview": "---\ntitle: Learning Rate Groups\ndescription: \"Setting different learning rates by module name\"\n---\n\n## Background\n\nInspi"
  },
  {
    "path": "docs/mac.qmd",
    "chars": 608,
    "preview": "---\ntitle: Mac M-series\ndescription: Mac M-series support\n---\n\nCurrently Axolotl on Mac is partially usable, many of the"
  },
  {
    "path": "docs/mixed_precision.qmd",
    "chars": 4855,
    "preview": "---\ntitle: \"Mixed Precision Training\"\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n    number-sections: true\n    code-"
  },
  {
    "path": "docs/multi-gpu.qmd",
    "chars": 5576,
    "preview": "---\ntitle: \"Multi-GPU\"\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n    # number-sections: true\n    code-tools: true\ne"
  },
  {
    "path": "docs/multi-node.qmd",
    "chars": 3263,
    "preview": "---\ntitle: Multi Node\ndescription: How to use Axolotl on multiple machines\n---\n\nThe below are three ways to train multi-"
  },
  {
    "path": "docs/multimodal.qmd",
    "chars": 7696,
    "preview": "---\ntitle: MultiModal / Vision Language Models (BETA)\nformat:\n  html:\n    toc: true\n    toc-depth: 3\n---\n\n## Supported M"
  },
  {
    "path": "docs/multipack.qmd",
    "chars": 2016,
    "preview": "---\ntitle: Multipack (Sample Packing)\ndescription: Multipack is a technique to pack multiple sequences into a single bat"
  },
  {
    "path": "docs/nccl.qmd",
    "chars": 3106,
    "preview": "---\ntitle: NCCL\ndescription: Troubleshooting NCCL issues\n---\n\nNVIDIA NCCL is a library to facilitate and optimize multi-"
  },
  {
    "path": "docs/nd_parallelism.qmd",
    "chars": 7587,
    "preview": "---\ntitle: \"N-D Parallelism (Beta)\"\n---\n\nAxolotl enables training models at scale by composing different parallelism tec"
  },
  {
    "path": "docs/optimizations.qmd",
    "chars": 6738,
    "preview": "---\ntitle: Optimizations Guide\ndescription: A guide to the performance and memory optimizations available in Axolotl.\n--"
  },
  {
    "path": "docs/optimizers.qmd",
    "chars": 3015,
    "preview": "---\ntitle: Optimizers\ndescription: Configuring optimizers\n---\n\n## Overview\n\nAxolotl supports all optimizers supported by"
  },
  {
    "path": "docs/qat.qmd",
    "chars": 2169,
    "preview": "---\ntitle: \"Quantization Aware Training (QAT)\"\nback-to-top-navigation: true\ntoc: true\ntoc-expand: 2\ntoc-depth: 4\n---\n\n##"
  },
  {
    "path": "docs/quantize.qmd",
    "chars": 2224,
    "preview": "---\ntitle: \"Quantization with torchao\"\nback-to-top-navigation: true\ntoc: true\ntoc-expand: 2\ntoc-depth: 4\n---\n\nQuantizati"
  },
  {
    "path": "docs/ray-integration.qmd",
    "chars": 3876,
    "preview": "---\ntitle: Ray Train\ndescription: How to use Axolotl with Ray Train\n---\n\nAxolotl supports using Ray as an alternative to"
  },
  {
    "path": "docs/reward_modelling.qmd",
    "chars": 2234,
    "preview": "---\ntitle: \"Reward Modelling\"\ndescription: \"Reward models are used to guide models towards behaviors which is preferred "
  },
  {
    "path": "docs/rlhf.qmd",
    "chars": 26077,
    "preview": "---\ntitle: \"RLHF (Beta)\"\ndescription: \"Reinforcement Learning from Human Feedback is a method whereby a language model i"
  },
  {
    "path": "docs/scripts/examples-allowlist.yml",
    "chars": 1487,
    "preview": "examples:\n  # December 2025\n  - name: kimi-linear\n    title: Kimi Linear\n  - name: plano\n    title: Plano Orchestrator\n "
  },
  {
    "path": "docs/scripts/generate_config_docs.py",
    "chars": 29346,
    "preview": "# type: ignore\n\n\"\"\"\nQuarto documentation generation from Pydantic models. Uses Pydantic model source code\nto automatical"
  },
  {
    "path": "docs/scripts/generate_examples_docs.py",
    "chars": 15598,
    "preview": "\"\"\"\nauto generate example docs from allowlist\n\"\"\"\n\nimport re\nimport shutil\nimport sys\nfrom pathlib import Path\n\nimport y"
  },
  {
    "path": "docs/sequence_parallelism.qmd",
    "chars": 3735,
    "preview": "---\ntitle: Sequence Parallelism\ndescription: Train with long sequences split across multiple GPUs.\n---\n\nSequence paralle"
  },
  {
    "path": "docs/streaming.qmd",
    "chars": 3425,
    "preview": "---\ntitle: Streaming Datasets\ndescription: How to use streaming mode for large-scale datasets and memory-efficient train"
  },
  {
    "path": "docs/telemetry.qmd",
    "chars": 2440,
    "preview": "---\ntitle: Telemetry\ndescription: A description of the telemetry implementation in Axolotl.\n---\n\n# Telemetry in Axolotl\n"
  },
  {
    "path": "docs/torchao.qmd",
    "chars": 731,
    "preview": "---\ntitle: \"PyTorch ao\"\ndescription: \"Custom data types and layouts for training and inference\"\n---\n\nTo use experimental"
  },
  {
    "path": "docs/unsloth.qmd",
    "chars": 1414,
    "preview": "---\ntitle: \"Unsloth\"\ndescription: \"Hyper-optimized QLoRA finetuning for single GPUs\"\n---\n\n### Overview\n\nUnsloth provides"
  },
  {
    "path": "examples/LiquidAI/README.md",
    "chars": 3153,
    "preview": "# Finetune Liquid Foundation Models 2 (LFM2) with Axolotl\n\n[Liquid Foundation Models 2 (LFM2)](https://huggingface.co/co"
  },
  {
    "path": "examples/LiquidAI/lfm2-350m-fft.yaml",
    "chars": 947,
    "preview": "base_model: LiquidAI/LFM2-350M\n\nplugins:\n  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin\n\neot_tokens:\n "
  },
  {
    "path": "examples/LiquidAI/lfm2-8b-a1b-lora.yaml",
    "chars": 1141,
    "preview": "base_model: LiquidAI/LFM2-8B-A1B\n\nplugins:\n  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin\n\nload_in_8bi"
  },
  {
    "path": "examples/LiquidAI/lfm2-vl-lora.yaml",
    "chars": 1278,
    "preview": "base_model: LiquidAI/LFM2-VL-450M\ntrust_remote_code: true\nmodel_type: AutoModelForImageTextToText\nprocessor_type: AutoPr"
  },
  {
    "path": "examples/alst/README.md",
    "chars": 1051,
    "preview": "# Arctic Long Sequence Training (ALST)\n\nArtic Long Sequence Training (ALST) is a technique for training long context mod"
  },
  {
    "path": "examples/alst/llama3-8b-deepspeed-alst.yaml",
    "chars": 1191,
    "preview": "base_model: meta-llama/Llama-3.1-8B\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/cus"
  },
  {
    "path": "examples/alst/llama3-8b-fsdp2-alst.yaml",
    "chars": 1416,
    "preview": "base_model: meta-llama/Llama-3.1-8B\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/cus"
  },
  {
    "path": "examples/apertus/README.md",
    "chars": 3838,
    "preview": "# Finetune Swiss-AI's Apertus with Axolotl\n\n[Apertus](https://huggingface.co/collections/swiss-ai/apertus-llm-68b699e654"
  },
  {
    "path": "examples/apertus/apertus-8b-qlora.yaml",
    "chars": 1161,
    "preview": "base_model: swiss-ai/Apertus-8B-Instruct-2509\n\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: u"
  },
  {
    "path": "examples/arcee/README.md",
    "chars": 2350,
    "preview": "# Finetune ArceeAI's AFM with Axolotl\n\n[Arcee Foundation Models (AFM)](https://huggingface.co/collections/arcee-ai/afm-4"
  },
  {
    "path": "examples/arcee/afm-4.5b-qlora.yaml",
    "chars": 1145,
    "preview": "base_model: arcee-ai/AFM-4.5B\n\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/custom_m"
  },
  {
    "path": "examples/archived/README.md",
    "chars": 195,
    "preview": "# Archived Examples\n\nThis directory contains examples that are no longer maintained and may no longer be functional.\n\nWe"
  },
  {
    "path": "examples/archived/cerebras/btlm-ft.yml",
    "chars": 1587,
    "preview": "base_model: cerebras/btlm-3b-8k-base\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCausal"
  },
  {
    "path": "examples/archived/cerebras/qlora.yml",
    "chars": 993,
    "preview": "base_model: cerebras/Cerebras-GPT-1.3B\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/"
  },
  {
    "path": "examples/archived/code-llama/13b/lora.yml",
    "chars": 1051,
    "preview": "base_model: codellama/CodeLlama-13b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM"
  },
  {
    "path": "examples/archived/code-llama/13b/qlora.yml",
    "chars": 1057,
    "preview": "base_model: codellama/CodeLlama-13b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM"
  },
  {
    "path": "examples/archived/code-llama/34b/lora.yml",
    "chars": 1051,
    "preview": "base_model: codellama/CodeLlama-34b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM"
  },
  {
    "path": "examples/archived/code-llama/34b/qlora.yml",
    "chars": 1057,
    "preview": "base_model: codellama/CodeLlama-34b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM"
  },
  {
    "path": "examples/archived/code-llama/7b/lora.yml",
    "chars": 1050,
    "preview": "base_model: codellama/CodeLlama-7b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\n"
  },
  {
    "path": "examples/archived/code-llama/7b/qlora.yml",
    "chars": 1056,
    "preview": "base_model: codellama/CodeLlama-7b-hf\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\n"
  },
  {
    "path": "examples/archived/code-llama/README.md",
    "chars": 769,
    "preview": "# Overview\n\nThis is an example of CodeLLaMA configuration for 7b, 13b and 34b.\n\nThe 7b variant fits on any 24GB VRAM GPU"
  },
  {
    "path": "examples/archived/dbrx/16bit-lora.yaml",
    "chars": 1510,
    "preview": "base_model: LnL-AI/dbrx-base-converted-v2\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: userna"
  },
  {
    "path": "examples/archived/dbrx/8bit-lora.yaml",
    "chars": 1550,
    "preview": "base_model: LnL-AI/dbrx-base-converted-v2\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: userna"
  },
  {
    "path": "examples/archived/dbrx/README.md",
    "chars": 1132,
    "preview": "# DBRX MoE\n\nCurrently, for LoRA, only the `q_proj`, `k_proj`, `v_proj` `out_proj` and `layer` Linear layers are trainabl"
  },
  {
    "path": "examples/archived/dbrx/fft-ds-zero3.yaml",
    "chars": 925,
    "preview": "base_model: LnL-AI/dbrx-base-converted-v2\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: userna"
  },
  {
    "path": "examples/archived/deepcoder/deepcoder-14B-preview-lora.yml",
    "chars": 941,
    "preview": "base_model: agentica-org/DeepCoder-14B-Preview\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: u"
  },
  {
    "path": "examples/archived/falcon/config-7b-lora.yml",
    "chars": 1264,
    "preview": "base_model: tiiuae/falcon-7b\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCausalLM\ntoken"
  },
  {
    "path": "examples/archived/falcon/config-7b-qlora.yml",
    "chars": 2193,
    "preview": "# 1b: tiiuae/falcon-rw-1b\n# 40b: tiiuae/falcon-40b\nbase_model: tiiuae/falcon-7b\n# optionally might have model_type or to"
  },
  {
    "path": "examples/archived/falcon/config-7b.yml",
    "chars": 1219,
    "preview": "base_model: tiiuae/falcon-7b\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCausalLM\ntoken"
  },
  {
    "path": "examples/archived/gemma/qlora.yml",
    "chars": 1028,
    "preview": "# use google/gemma-7b if you have access\nbase_model: mhenrichsen/gemma-7b\n# optionally might have model_type or tokenize"
  },
  {
    "path": "examples/archived/gptj/qlora.yml",
    "chars": 980,
    "preview": "base_model: EleutherAI/gpt-j-6b\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/custom_"
  },
  {
    "path": "examples/archived/jeopardy-bot/config.yml",
    "chars": 1069,
    "preview": "base_model: huggyllama/llama-7b\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\ntokeni"
  },
  {
    "path": "examples/archived/mpt-7b/README.md",
    "chars": 89,
    "preview": "# MPT-7B\n\n```shell\naccelerate launch scripts/finetune.py examples/mpt-7b/config.yml\n\n```\n"
  },
  {
    "path": "examples/archived/mpt-7b/config.yml",
    "chars": 1200,
    "preview": "base_model: mosaicml/mpt-7b\n# optionally might have model_type or tokenizer_type\ntokenizer_type: AutoTokenizer\n# Automat"
  },
  {
    "path": "examples/archived/openllama-3b/README.md",
    "chars": 294,
    "preview": "# openllama-3b\n\nBasic full tune\n```shell\naccelerate launch scripts/finetune.py examples/openllama-3b/config.yml\n```\n\nLoR"
  },
  {
    "path": "examples/archived/openllama-3b/config.yml",
    "chars": 1105,
    "preview": "base_model: openlm-research/open_llama_3b_v2\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCa"
  },
  {
    "path": "examples/archived/openllama-3b/lora.yml",
    "chars": 1201,
    "preview": "base_model: openlm-research/open_llama_3b_v2\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCa"
  },
  {
    "path": "examples/archived/openllama-3b/qlora.yml",
    "chars": 1127,
    "preview": "base_model: openlm-research/open_llama_3b_v2\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCa"
  },
  {
    "path": "examples/archived/pythia/lora.yml",
    "chars": 778,
    "preview": "base_model: EleutherAI/pythia-1.4b-deduped\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: usern"
  },
  {
    "path": "examples/archived/pythia-12b/README.md",
    "chars": 197,
    "preview": "# Pythia 12B\n\n- Single-GPU A100 only (?)\n\n```shell\npython scripts/finetune.py examples/pythia-12b/config.yml\n```\n\n⚠️ Mul"
  },
  {
    "path": "examples/archived/pythia-12b/config.yml",
    "chars": 1005,
    "preview": "base_model: EleutherAI/pythia-12b-deduped\nbase_model_ignore_patterns: pytorch*  # prefer safetensors\n# optionally might "
  },
  {
    "path": "examples/archived/qwen/README.md",
    "chars": 110,
    "preview": "# Qwen\n\nTODO\n\n# Qwen2 MoE\n\n✅ multipack\n✅ qwen2_moe 4-bit QLoRA\n✅ qwen2_moe 16-bit LoRA\n❓ qwen2_moe 8-bit LoRA\n"
  },
  {
    "path": "examples/archived/qwen/lora.yml",
    "chars": 1041,
    "preview": "base_model: Qwen/Qwen-7B\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCausalLM\ntokenizer"
  },
  {
    "path": "examples/archived/qwen/qlora.yml",
    "chars": 1042,
    "preview": "base_model: Qwen/Qwen-7B\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCausalLM\ntokenizer"
  },
  {
    "path": "examples/archived/qwen/qwen2-moe-lora.yaml",
    "chars": 954,
    "preview": "base_model: Qwen/Qwen1.5-MoE-A2.7B\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/cust"
  },
  {
    "path": "examples/archived/qwen/qwen2-moe-qlora.yaml",
    "chars": 995,
    "preview": "base_model: Qwen/Qwen1.5-MoE-A2.7B\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/cust"
  },
  {
    "path": "examples/archived/redpajama/README.md",
    "chars": 117,
    "preview": "# RedPajama 3B preview release\n\n```shell\naccelerate launch scripts/finetune.py examples/redpajama/config-3b.yml\n\n```\n"
  },
  {
    "path": "examples/archived/redpajama/config-3b.yml",
    "chars": 1173,
    "preview": "base_model: togethercomputer/RedPajama-INCITE-Chat-3B-v1\n# optionally might have model_type or tokenizer_type\nmodel_type"
  },
  {
    "path": "examples/archived/replit-3b/config-lora.yml",
    "chars": 907,
    "preview": "base_model: replit/replit-code-v1-3b\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/cu"
  },
  {
    "path": "examples/archived/stablelm-2/1.6b/fft.yml",
    "chars": 1111,
    "preview": "base_model: stabilityai/stablelm-2-1_6b\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCau"
  },
  {
    "path": "examples/archived/stablelm-2/1.6b/lora.yml",
    "chars": 1074,
    "preview": "base_model: stabilityai/stablelm-2-1_6b\n# optionally might have model_type or tokenizer_type\nmodel_type: AutoModelForCau"
  },
  {
    "path": "examples/archived/stablelm-2/README.md",
    "chars": 1577,
    "preview": "# StableLM 2\n\nThis repository contains examples for training and processing using StableLM-2. It also includes a section"
  },
  {
    "path": "examples/archived/starcoder2/qlora.yml",
    "chars": 921,
    "preview": "base_model: bigcode/starcoder2-3b\n# Automatically upload checkpoint and final model to HF\n# hub_model_id: username/custo"
  },
  {
    "path": "examples/archived/tiny-llama/README.md",
    "chars": 319,
    "preview": "# Overview\n\nThis is a simple example of how to finetune TinyLlama1.1B using either lora or qlora:\n\nLoRa:\n\n```\naccelerate"
  },
  {
    "path": "examples/archived/tiny-llama/lora-mps.yml",
    "chars": 1024,
    "preview": "base_model: TinyLlama/TinyLlama_v1.1\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\nt"
  },
  {
    "path": "examples/archived/tiny-llama/lora.yml",
    "chars": 982,
    "preview": "base_model: TinyLlama/TinyLlama_v1.1\n# optionally might have model_type or tokenizer_type\ntokenizer_type: AutoTokenizer\n"
  },
  {
    "path": "examples/archived/tiny-llama/pretrain.yml",
    "chars": 876,
    "preview": "base_model: TinyLlama/TinyLlama-1.1B-Chat-v1.0\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaFor"
  },
  {
    "path": "examples/archived/tiny-llama/qlora.yml",
    "chars": 1018,
    "preview": "base_model: TinyLlama/TinyLlama_v1.1\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\nt"
  },
  {
    "path": "examples/archived/xgen-7b/xgen-7b-8k-qlora.yml",
    "chars": 2200,
    "preview": "# An example finetuning Saleforce's XGen-7b model with 8k context using qlora\n# on Tim Dettmer's Guanaco dataset.\nbase_m"
  },
  {
    "path": "examples/archived/yi-34B-chat/README.md",
    "chars": 348,
    "preview": "# Overview\n\nThis is an example of a Yi-34B-Chat configuration. It demonstrates that it is possible to finetune a 34B mod"
  },
  {
    "path": "examples/archived/yi-34B-chat/qlora.yml",
    "chars": 1146,
    "preview": "base_model: 01-ai/Yi-34B-Chat\n# optionally might have model_type or tokenizer_type\nmodel_type: LlamaForCausalLM\ntokenize"
  },
  {
    "path": "examples/cloud/baseten.yaml",
    "chars": 111,
    "preview": "provider: baseten\nproject_name:\n\nsecrets:\n  - HF_TOKEN\n  - WANDB_API_KEY\n\ngpu: h100\ngpu_count: 8\nnode_count: 1\n"
  },
  {
    "path": "examples/cloud/modal.yaml",
    "chars": 523,
    "preview": "project_name:\nvolumes:\n  - name: axolotl-data\n    mount: /workspace/data\n  - name: axolotl-artifacts\n    mount: /workspa"
  }
]

// ... and 870 more files (download for full content)